Preliminary Test of a Data Assimilation System with a Regional High-Resolution Atmosphere–Ocean Coupled Model Based on an Ensemble Kalman Filter

Masaru Kunii Forecast Research Department, Meteorological Research Institute, Tsukuba, Japan

Search for other papers by Masaru Kunii in
Current site
Google Scholar
PubMed
Close
,
Kosuke Ito Forecast Research Department, Meteorological Research Institute, Tsukuba, and Department of Physics and Geosciences, University of the Ryukyus, Okinawa, Japan

Search for other papers by Kosuke Ito in
Current site
Google Scholar
PubMed
Close
, and
Akiyoshi Wada Typhoon Research Department, Meteorological Research Institute, Tsukuba, Japan

Search for other papers by Akiyoshi Wada in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

An ensemble Kalman filter (EnKF) that uses a regional mesoscale atmosphere–ocean coupled model was preliminarily examined to provide realistic sea surface temperature (SST) estimates and to represent the uncertainties of SST in ensemble data assimilation strategies. The system was evaluated through data assimilation cycle experiments over a one-month period from July to August 2014, during which time a tropical cyclone (TC) as well as severe rainfall events occurred. The results showed that the data assimilation cycle with the coupled model reproduced SST distributions realistically even without assimilating SST and sea surface salinity observations, and atmospheric variables provided to ocean models can, therefore, control oceanic variables physically to some extent. The forecast error covariance calculated in the EnKF with the coupled model showed dependency on oceanic vertical mixing for near-surface atmospheric variables due to the difference of variability between the atmosphere and the ocean as well as the influence of SST variations on the atmospheric boundary layer. The EnKF with the coupled model reproduced the intensity change of Typhoon Halong (2014) during the mature phase more realistically than with an uncoupled atmosphere model, although there remained a degradation of the SST estimate, particularly around the Kuroshio region. This suggests that an atmosphere–ocean coupled data assimilation system should be developed that is able to physically control both atmospheric and oceanic variables.

© 2017 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author e-mail: Masaru Kunii, mkunii@mri-jma.go.jp

Abstract

An ensemble Kalman filter (EnKF) that uses a regional mesoscale atmosphere–ocean coupled model was preliminarily examined to provide realistic sea surface temperature (SST) estimates and to represent the uncertainties of SST in ensemble data assimilation strategies. The system was evaluated through data assimilation cycle experiments over a one-month period from July to August 2014, during which time a tropical cyclone (TC) as well as severe rainfall events occurred. The results showed that the data assimilation cycle with the coupled model reproduced SST distributions realistically even without assimilating SST and sea surface salinity observations, and atmospheric variables provided to ocean models can, therefore, control oceanic variables physically to some extent. The forecast error covariance calculated in the EnKF with the coupled model showed dependency on oceanic vertical mixing for near-surface atmospheric variables due to the difference of variability between the atmosphere and the ocean as well as the influence of SST variations on the atmospheric boundary layer. The EnKF with the coupled model reproduced the intensity change of Typhoon Halong (2014) during the mature phase more realistically than with an uncoupled atmosphere model, although there remained a degradation of the SST estimate, particularly around the Kuroshio region. This suggests that an atmosphere–ocean coupled data assimilation system should be developed that is able to physically control both atmospheric and oceanic variables.

© 2017 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author e-mail: Masaru Kunii, mkunii@mri-jma.go.jp

1. Introduction

An ensemble Kalman filter (EnKF) is an advanced data assimilation technique that has been widely used for data assimilation studies since it was first proposed by Evensen (1994). Compared to other data assimilation techniques, such as the four-dimensional variational data assimilation system (e.g., Lewis and Derber 1985; Courtier et al. 1994), EnKF has an advantage in that it can estimate atmospheric flow-dependent background error covariance spontaneously from ensemble perturbations. This advantage is useful for simulations of extreme weather events, where consideration of atmospheric flow-dependent background errors becomes essential (Hamill et al. 2011). In addition, the practical application of EnKF is easy, although special treatments are required to deal with sampling noise and underestimation of the background error covariance, which are mainly caused by insufficient representation of model errors as well as the approximation of the original Kalman filter with a finite number of ensemble members. Moreover, in EnKF it is not necessary to use tangent-linear and adjoint versions of forecast models.

After promising results in the application of EnKF to global numerical weather prediction (NWP) systems, the technique has also been applied to a regional NWP model (e.g., Snyder and Zhang 2003; Dowell et al. 2004; Bonavita et al. 2008; Miyoshi and Kunii 2012; Kunii and Miyoshi 2012; Kunii 2014a). For its application to a limited-area mesoscale model such as a regional NWP system, the treatment of boundary condition uncertainties has been a major issue. When identical boundary conditions are equally applied in ensemble members for regional ensemble forecasts, uncertainties near the boundaries are not taken into consideration, the result being universal overconfidence of first-guess forecasts. In the above-mentioned previous studies, identical lateral boundary conditions were used for all ensemble members but with a large enough domain to minimize the effect of small perturbations that occur near the boundaries. On the other hand, there are also numerous studies that have addressed uncertainties of lateral boundaries by using lateral boundary perturbations derived from global ensemble systems (Hou et al. 2001; Bowler and Mylne 2009; Kunii 2014a), or provided by random draws from spatial and temporal covariance relationships (Barker 2005; Torn et al. 2006; Schwartz et al. 2015). Recently, the differences between these perturbing strategies were systematically studied by Ouaraini et al. (2015).

Kunii and Miyoshi (2012) have shown the advantage of including sea surface temperature (SST) uncertainties for the lowermost boundary over the ocean by using climatological perturbations in local ensemble transform Kalman filter (LETKF) data assimilation cycles. The ensemble spreads for air temperature and moisture at lower levels increased with SST perturbations. The analysis fields improved, as did subsequent model forecasts for a tropical cyclone (TC) event, whereas the manually inflated ensemble spread did not contribute to the improvements. McLay et al. (2012) reported that ensemble-transform-based SST perturbations have a broader benefit in the tropics than SST variations calculated from the prognostic diurnal SST scheme. However, SST variations are generally determined through complicated dynamic and thermodynamic processes, such as advection and mixing (e.g., Kawai and Wada 2007). For these reasons, further studies of more sophisticated procedures for producing SST perturbations are needed. In fact, the SST uncertainties employed by Kunii and Miyoshi (2012) were primitive; that is, fixed SST perturbations were applied throughout the data assimilation cycles, and thus the evolution of SST uncertainties and the variation were not taken into consideration.

To assess SST time-variable uncertainties in an EnKF, its application with an atmosphere–ocean coupled model is an appropriate choice. Implementing an atmosphere–ocean coupled model in an EnKF can also be expected to provide more realistic SST estimates, which would be particularly suitable for TC cases. Although an EnKF with a coupled model has been applied for climate simulations (e.g., Liu et al. 2014; Pendergrass et al. 2012; Tardif et al. 2014; Wu et al. 2012; Zhang et al. 2007, 2012), the application of coupled data assimilation is still in the early stages for mesoscale models that aim at short-range NWP forecasts (Wada and Kunii 2014), despite the fact that a great deal of research has been carried out on mesoscale forecasts that use atmosphere–ocean coupled models, mainly for TCs (e.g., Chen et al. 2013; Wada et al. 2014; Ito et al. 2015). Ito et al. (2015) developed a high-resolution atmosphere–ocean coupled model by coupling a vertically one-dimensional upper ocean model (Price et al. 1986, hereafter the PWP model) with the Japan Meteorological Agency (JMA) nonhydrostatic model (NHM; Saito et al. 2006, 2007; Saito 2012). To evaluate the impact on TC forecasts statistically, they applied the coupled NHM (CNHM)1 to 34 TCs that passed in the vicinity of Japan from 2009 and 2012. The results showed the superiority of the CNHM over the original NHM, especially at later forecast times. The improvement rates eventually exceeded 20% for minimum sea level pressure at the 48-h forecast time.

In the current study, SST uncertainties were directly evaluated by using the CNHM in an EnKF. The utilization of a regional high-resolution atmosphere–ocean coupled model in an EnKF should lead to more appropriate estimation of SST and its uncertainties based on physics, and thus it will increase the accuracy of analyses and subsequent model forecasts, particularly for events in which SST plays an important role, such as TCs. The present study aims to clarify the sensitivity of the atmospheric analysis to physically based SST estimation and its perturbations calculated during an EnKF data assimilation cycle with a regional high-resolution atmosphere–ocean coupled model. The remainder of the article is organized as follows. Section 2 describes the experimental design and settings. The results of the experiments are presented in section 3. Finally, a summary and discussion are provided in section 4.

2. Experimental design and settings

a. NHM-LETKF system

An LETKF (Hunt et al. 2007) system implemented with an NHM (NHM-LETKF; Kunii 2014a) was used in this study. The system provided promising results with the analyses and subsequent model forecasts for a local severe rainfall event in 2012 in Japan (Kunii 2014a). Since then, the NHM-LETKF system has been implemented on the K computer, the flagship supercomputer in Japan. The performance of the system on the K computer as well as the influence of sampling noise on the background error covariance were evaluated with 1000 ensemble members (Kunii 2014b). In addition, the system was used with a chemical transport model to improve analyses and simulations of the Fukushima nuclear accident on 15 March 2011 (Sekiyama et al. 2015).

This study applied configurations similar to those used by Kunii (2014a). The computational domain in the system was collocated to cover all of Japan and the surrounding area (Fig. 1). The number of grid points was 273 × 221 in the Lambert conformal projection, with a horizontal grid spacing of 15 km and 50 vertical levels. For the physical processes in the NHM, the modified Kain–Fritsch convective parameterization scheme (Ohmori and Yamada 2006) was used in conjunction with three-ice bulk cloud microphysics (Ikawa and Saito 1991). A Mellor–Yamada level-3 closure model (Nakanishi and Niino 2004, 2006) with a partial condensation scheme was used as a turbulence scheme.

Fig. 1.
Fig. 1.

Model domain for the NHM-LETKF analysis and the subsequent model forecast. Colors indicate land elevation in meters.

Citation: Monthly Weather Review 145, 2; 10.1175/MWR-D-16-0068.1

Lateral boundary conditions for running the NHM were provided from the outputs of the JMA operational global forecasts. To prevent underestimation of ensemble spreads near lateral boundaries, lateral boundary uncertainties were included by adding perturbations derived from the JMA operational one-week ensemble prediction system, which follows Saito et al. (2012). Kunii (2014a) have validated the effectiveness of including lateral boundary perturbations.

The atmospheric variables controlled in the LETKF system were three-dimensional wind components (u, υ, w), temperature T, pressure p, water vapor mixing ratio qυ, and water/ice microphysics variables. The covariance localization scales used were set to be 200 km in the horizontal and 0.2 in the vertical in the logp coordinate system, and 3 h in time. The localization parameters corresponded to the 1σ length, at which point the Gaussian localization function became e−0.5. The assimilation window length was 6 h, and observation data rounded to the nearest hour were assimilated six times in the window. Assimilated observation data included radiosonde, pilot balloon, wind profiler, aircraft, ship, buoy, and total precipitable water data obtained from the Global Navigation Satellite System. Surface wind and atmospheric motion vectors over the ocean derived from satellites were also used. Other satellite data, such as those from the Special Sensor Microwave Imager, Tropical Rainfall Measuring Mission Microwave Imager, and Advanced Microwave Scanning Radiometer 2, were not assimilated in this study, even though the data are assimilated in the operational mesoscale data assimilation system of JMA. To estimate inflating background error covariance, the relaxation to prior spread method proposed by Whitaker and Hamill (2012) was adopted. This is a kind of multiplicative inflation method, in which the inflation parameter is determined when the analyzed ensemble standard deviation is relaxed to the prior estimate. We set the relative weight of the background (analysis) to 0.95 (0.05) as in Whitaker and Hamill (2012).

b. Data assimilation cycles with the atmosphere–ocean coupled model

CNHM (Ito et al. 2015) was newly implemented as a forecast model in the NHM-LETKF system. As described in the introduction, the oceanic part of CNHM is the PWP model. PWP is a very simple mixed layer model that can represent the diurnal cycle and vertical structure of the upper ocean at a fixed grid, but it is not a general circulation model. It calculates ocean temperature, salinity, and horizontal components of current velocity based on the inputs of longwave and shortwave radiation, sensible and latent heat fluxes, and wind stress calculated in the NHM. Vertical mixing is calculated by a combination of stability criteria for static stability, mixed layer shear flow stability, and stratified shear flow stability based on density stratification, bulk Richardson number, and gradient Richardson number (Price et al. 1994).

Because PWP is a convenient ocean model, CNHM has limitations in that three-dimensional processes such as horizontal advection, particularly around the coastal region, mixing and upwelling induced by a TC, and inertial-gravity wave propagation to the ocean interior cannot be reproduced realistically. The vertical grid spacing of the ocean model was 5 m, with the deepest bottom at a depth of 400 m (or the bottom of the ocean if it was between 50 and 400 m) for simplicity and to save computational time. The initial states of the oceanic variables were created following Ito et al. (2015). The coupling process was carried out between the atmosphere and the ocean every 10 min. The configuration of CNHM consumed no more than approximately 1% of the total computational time of the original NHM.

To clarify the impact of the EnKF system with CNHM on analyses and subsequent model forecasts, the following two data assimilation experiments were conducted. The “CTL” experiment used the original settings of the NHM-LETKF system, that is, the original NHM with the Merged Satellite and In situ Data Global Daily Sea Surface Temperatures (MGDSST) dataset produced by JMA (Kurihara et al. 2006; JMA 2013) as the lowermost boundary over the ocean. The “CPL” experiment adopted CNHM instead of the original NHM to generate the first-guess outputs during the data assimilation cycles. For simplicity, the oceanic variables calculated by CNHM were propagated in the subsequent cycles without being updated with observations through the data assimilation process. This means that the oceanic variables were not treated as control variables in the LETKF. The procedure was the same as the other lower boundary conditions such as ground temperature and volumetric soil water content. Although the current configuration leads to well-balanced background fields for the atmospheric and ocean variables, the analyses could potentially become imbalanced because only atmospheric variables are updated in the data assimilation procedure. The problem would be solved by consideration of a fully coupled data assimilation by simultaneous treatment of atmospheric and oceanic error covariance in the EnKF system. This will be an important subject for future research.

c. Data assimilation experiments

The NHM-LETKF system with CNHM was evaluated through data assimilation and subsequent model forecast experiments during the summer months of 2014. It was carried out from 0000 UTC 10 July to 0000 UTC 20 August 2014. The experimental period included not only severe local rainfall events throughout all of Japan but also TC Halong. A tropical depression was upgraded to TC Halong at 0000 UTC 29 July near the Mariana Islands. It moved west-northwestward with rapid intensification from 1200 UTC 1 August to 1200 UTC 2 August according to best track data from the Regional Specialized Meteorological Center Tokyo. After the minimum sea level pressure reached 920 hPa, Halong moved northward with an average translation speed of approximately 4 m s−1. It made landfall on Shikoku Island in Japan at around 2100 UTC 9 August. After 0000 UTC 11 August, Halong became an extratropical cyclone. It caused torrential rainfalls while passing over the Shikoku and Kinki districts. The observed accumulated rainfall from 30 July to 11 August 2014 exceeded 1000 mm at some observational stations on Shikoku Island.

After the data assimilation experiments, single deterministic NHM forecasts were carried out from 0000 UTC 7 August to 1800 UTC 19 August 2014 every 6 h by using the ensemble-mean analyses based on the results of LETKF as the initial conditions. The model configurations for the forecasts were similar to those in the data assimilation cycles, but they had a longer lead time (up to 48 h) and a higher horizontal resolution (5 km). To examine the influence of the initial atmospheric and oceanic states on the forecast separately, three extended forecast runs were executed. In the “CTLM” experiment, the atmospheric states were derived from the CTL data assimilation experiments, and MGDSST was used for SST estimates as a lower boundary condition in the CTL data assimilation cycles.

Forecast experiments initialized with the CPL analyses were divided into two experiments to examine the isolated impacts of the atmospheric and oceanic initial conditions because the data assimilation with CNHM provided both atmospheric and oceanic states. The initial atmospheric states in the “CPLM” experiment came from the CPL analyses, and MGDSST was adopted instead of the ensemble-mean SST analyzed in the CPL experiment. The “CPLO” experiment was initialized with the CPL data assimilation results for both the atmospheric and oceanic states. In the present study, we did not conduct another “CTLO” experiment initialized with the atmospheric state from the CTL analysis and the oceanic state from the CPL analysis. Such an experiment would be unrealistic because the atmosphere tends to respond to the ocean on a much faster time scale than the ocean responds to the atmosphere. In these single 48-h deterministic NHM forecast experiments, the uncoupled NHM was used to examine the isolated impacts of initial conditions. The specifications of the forecast experiment settings are summarized in Table 1.

Table 1.

Summary of the single deterministic forecast experiments.

Table 1.

3. Results

a. Critical bulk Richardson number

Because the SST fields were not updated during a data assimilation cycle in the configuration of this study, one of the major concerns in regard to system stability was the SST bias that would occur through a long data assimilation period. In principle, information about observational data would have indirect influences even on noncontrol variables through the interactions between noncontrol and updated control variables. However, it is possible that calculations by CNHM cause an SST bias during a longer data assimilation period partly because of imperfect physical parameterization schemes in the ocean model (e.g., Zhang et al. 2012) and uncertainties of atmospheric forcing attributed to imperfect physical schemes in the atmosphere model. In addition, observation data for atmospheric variables over the ocean were relatively scarce, so that the SST bias is difficult to correct once it is generated.

Figure 2a shows the SST time series averaged over the computational domain. In the domain, the SST fields calculated by CNHM were changed through the atmosphere–ocean interaction by using CNHM. Owing to the application of identical oceanic variables for all ensemble members in the first cycle, the SST perturbations were small at the beginning of the data assimilation cycle. As the cycle proceeded, the ensemble spread of SST increased gradually and fluctuated around 0.18 K after about 10 days. At that time, the difference between averaged SST from the CPL experiment and that from MGDSST became obvious. The maximum amplitude of the negative bias of SST in the CPL experiment eventually reached approximately 2 K during the experimental period.

Fig. 2.
Fig. 2.

Time series of SST averaged over the model domain obtained from daily mean SST (MGDSST), ensemble mean SST in the CPL experiment (CPL), daily mean SST (OISST) based on satellite observations (OBS), and the ensemble spread of SST in the CPL experiment (Spread). The critical bulk Richardson numbers were (a) 0.6 and (b) 0.25.

Citation: Monthly Weather Review 145, 2; 10.1175/MWR-D-16-0068.1

Because the negative bias of SST would lead to negative impacts on air temperature and relative humidity near the surface and in the atmospheric boundary layers, and finally on precipitation forecasts, we first tried to fix this negative feedback by restoring the SST to MGDSST for each step of the exchange process in the CNHM between the atmosphere and the ocean. Although this procedure reduced the negative bias as expected, it simultaneously suppressed the growth of the ensemble spread of SST, resulting in a reduction in SST uncertainties because in all ensemble members the SSTs were equally nudged to the MGDSST with this restoration procedure.

We next examined the impact of the setting of the critical bulk Richardson number on mitigation of the SST bias induced by the ocean model in the CPL experiment. The critical bulk Richardson number Rb is a parameter related to vertical mixing for mixed layer shear flow stability in the ocean model and is defined as
e1
where g (m s−2) is the acceleration of gravity, (kg m−3) is the density of seawater, h (m) is the oceanic mixed layer depth, V (m s−1) is the current velocity, and kg m−3; determines the difference between the oceanic mixed layer and the level just beneath it. The parameter Rb is defined by two metrics such as mixed layer depth and stratification in addition to vertical shear in the mixed layer [Eq. (1)]. The parameter Rb in CNHM was originally set at 0.6, following Ito et al. (2015), which is different from the value of 0.65 used in the PWP model. The result of Ito et al. (2015) showed that the parameter value of 0.6 did not cause any negative SST biases on short-range TC forecasts up to 72 h. However, in the current experiments, the negative SST bias became serious on relatively long-range forecasts of more than one month, so the parameter value needed to be tuned.

Figure 2b shows the same time series as that shown in Fig. 2a, but with the parameter Rb of 0.25 in the CPL experiment. The lower value of the parameter inhibited excessive vertical mixing in the upper ocean and resultant decrease in SST while maintaining the ensemble spread. In addition, the decrease in SST induced by TC Halong was well captured around 10 August 2014, which is consistent with the microwave optimally interpolated SST product (OISST) by Remote Sensing Systems (http://www.remss.com/measurements/sea-surface-temperature/oisst-description).

Under the assumption of constant vertical shear in the mixed layer, the smaller Rb would lead to the amelioration of the negative bias of SST due to suppression of mixed layer deepening. However, this treatment may lead to a negative effect on ocean variables other than SST. Figure 3 depicts vertical cross sections of the differences in ocean temperature, salinity, and zonal current velocity between the CPL experiments with parameters of 0.25 and 0.6 as a function of depth and longitude at 30°N at 0000 UTC 10 August 2014. The impact of the tuning parameter on the below-surface ocean variables was limited to the upper levels, including the mixed layer, and it was relatively small for ocean salinity and zonal current velocity.

Fig. 3.
Fig. 3.

Vertical cross sections of the differences in ocean (a) temperature (K), (b) salinity, and (c) zonal current velocity (m s−1) between the CPL experiments with critical bulk Richardson numbers of 0.25 and 0.6 as a function of depth (m) and longitude at 30°N at 0000 UTC 10 Aug 2014. The contour interval is 0.2.

Citation: Monthly Weather Review 145, 2; 10.1175/MWR-D-16-0068.1

The reduction in SST bias due to the small tuning parameter 0.25 implies that the oceanic environment, including the oceanic initial conditions based on oceanic climatological data, is favorable for vertical mixing in the upper ocean and resultant sea surface cooling. In other words, the lower value of Rb is simply compensating for deficiencies of the oceanic initial condition and a simple vertically one-dimensional ocean model as well as uncertainties in the wind forcing. However, the impact of upper-ocean stratification on the SST bias significantly grows as the forecast time becomes long. In addition, it should be noted that atmospheric responses such as wind stresses, radiation, and cloud properties to the tuning parameter 0.25 and resultant SST variations were well balanced with the atmospheric analysis. Therefore, based on the result shown in Fig. 3, we use a parameter of 0.25 in the following for the test of the data assimilation system with CNHM.

b. Verification of the SST field

Figure 4 shows horizontal distributions of the daily mean SST obtained from MGDSST and OISST, and the ensemble mean SST in the CPL experiment on 10 August 2014. At that time, Halong made landfall on Shikoku Island in Japan. The center position was located near 34.2°N, 134.3°E. The daily mean SST obtained from MGDSST showed reasonable agreement with that obtained from OISST in most of the computational domain, whereas it failed to capture the rapid decrease in SST induced by Halong (Fig. 4a). In contrast, the Halong-induced cold wake that occurred along Halong’s path was better captured in the CPL experiment (Fig. 4b). However, the comparison between the OISST and the ensemble mean SST in the CPL experiment (Figs. 4b and 4c) showed that the SST in the CPL experiment had positive (negative) bias in high (low) latitudes in the domain. Figures 5a and 5b depict the horizontal distributions of root-mean-square errors (RMSEs) of SSTs from the MGDSST (Fig. 5a) and the CPL experiment (Fig. 5b) relative to the OISST averaged over 19 days from 0000 UTC 1 August to 1800 UTC 19 August 2014, respectively. As can be seen in Fig. 4a, the RMSE of MGDSST showed remarkably large errors along Halong’s path (Fig. 5a). These large errors likely correspond to the large bias in MGDSST observed around the location of Halong when it made landfall in Japan (Fig. 5c). Although the error due primarily to the cold wake induced by Halong was improved by using CNHM in the CPL experiment, relatively large RMSEs near the coast and those in high latitude became obvious (Fig. 5b). In addition, the negative bias of SST in the CPL experiment was excessive after the passage of Halong (Fig. 5c).

Fig. 4.
Fig. 4.

Horizontal distributions of (a) daily mean SST obtained from the MGDSST, (b) the ensemble mean SST in the CPL experiment, and (c) daily mean SST obtained from the OISST on 10 Aug 2014. Shades and contours indicate the value of SST. The contour interval is 1°C.

Citation: Monthly Weather Review 145, 2; 10.1175/MWR-D-16-0068.1

Fig. 5.
Fig. 5.

Horizontal distributions of RMSEs of (a) daily mean SST obtained from the MGDSST and (b) the ensemble mean SST in the CPL experiment relative to OISST, averaged over 19 days from 0000 UTC 1 Aug to 1800 UTC 19 Aug 2014. (c) Time series of bias and RMSE of the daily mean SST obtained from the MGDSST (black) and the ensemble mean SST in the CPL experiment (red) relative to OISST, averaged over the region 25°–40°N, 125°–145°E.

Citation: Monthly Weather Review 145, 2; 10.1175/MWR-D-16-0068.1

Figures 2 and 4 show that the horizontal distribution of SST based on MGDSST was markedly different from that based on OISST, despite the fact that they had the same horizontal resolution of 0.25°. OISST tended to better capture local changes in the SST distribution even though the distribution seems to be noisy because in MGDSST a period of SST variations shorter than 27 days is cut off to suppress the high-frequency SST noise embedded in satellite observations (JMA 2013), whereas those are included in OISST. In fact, the time scale of the oceanic response to a TC and the recovery to the normal condition (Price et al. 2008; Dare and McBride 2011) is usually shorter than the cut-off frequency applied to MGDSST, so that the error of MGDSST relative to OISST became large along the pathway of Halong. The large RMSE for the SST in the CPL experiment seems to be attributable to the limitation of the upper ocean model used in this study. For example, the ocean model used in this study cannot reproduce the effect of tidal mixing, which became dominant for SST variations near the coastal region. In addition, the error of the ocean model tended to be large over strong currents such as the Kuroshio and the Oyashio Current. Although the SSTs over the strong currents were significantly affected by three-dimensional oceanic processes, these effects were not included in the present configuration. In fact, Fig. 5b indicates that the SST estimates had relatively large errors over the Kuroshio.

c. Investigation of the analyzed fields

This subsection describes the impact on the analyzed atmospheric fields of the inclusion of CNHM in the data assimilation system based on EnKF. Figure 6 shows vertical profiles of the ensemble spread averaged over the computational domain for 76 samples from 0000 UTC 1 August to 1800 UTC 19 August every 6 h. The spinup period around the time of initiation of the data assimilation cycle was not included in the analysis. By including the SST uncertainties calculated by CNHM every LETKF cycle, relatively large ensemble spreads were obtained for lower-level air temperature and water vapor (Figs. 6b and 6c), whereas the impact was negligible for the horizontal winds (Fig. 6a). The difference between the vertical profiles in the CTL and CPL experiments was particularly noticeable for the air temperature field at lower model levels. The ensemble spread for water vapor in the CPL experiment increased slightly around 950 hPa (eighth model level), but decreased around the 850-hPa level (13th model level) compared to that in the CTL experiment.

Fig. 6.
Fig. 6.

Vertical profiles of the ensemble spread for (a) meridional wind component (m s−1), (b) temperature (K), and (c) water vapor mixing ratio (g kg−1), averaged over the model domain for 76 samples from 0000 UTC 1 Aug to 1800 UTC 19 Aug 2014 every 6 h. Blue dashed lines indicate the results from the CTL experiment. Red solid lines indicate the results from the CPL experiment.

Citation: Monthly Weather Review 145, 2; 10.1175/MWR-D-16-0068.1

The variation in the ensemble spread was related to the correlation between SST and atmospheric variables. Figure 7 illustrates vertical cross sections of the error correlation between SST and the meridional wind, air temperature, water vapor, and vertical wind fields. As seen in Fig. 6a, there is no notable correlation between SST and horizontal wind (Fig. 7a). However, SST and air temperature appear to be strongly correlated, particularly below the 950-hPa level (Fig. 7b), corresponding to the atmospheric boundary layer. The correlations are a manifestation of the effect on the atmospheric boundary layer through the atmosphere–ocean coupled processes in CNHM. The correlation between SST and water vapor shows small positive impacts around the 950-hPa level but turns negative at higher levels than the 950-hPa level, corresponding to the results shown in Fig. 6c. For the vertical wind, a small positive correlation can be seen around the 975-hPa level (Fig. 7d). It can be assumed that this positive correlation originates from the positive vertical wind anomaly induced by local instability near the sea surface. As a consequence, the vertical wind anomaly would contribute to the vertical transportation of water vapor, the result being the positive correlation and relatively large ensemble spread of water vapor around the 950-hPa level (Figs. 7c and 6c).

Fig. 7.
Fig. 7.

Vertical cross sections of the error correlation between the SST from the domain-center location (denoted by the cross marks, 32.5°N, 140.0°E) and (a) meridional wind, (b) air temperature, (c) water vapor mixing ratio, and (d) vertical wind. The correlation was an average of 76 samples from 0000 UTC 1 Aug to 1800 UTC 19 Aug 2014 every 6 h. Contour intervals are 0.1.

Citation: Monthly Weather Review 145, 2; 10.1175/MWR-D-16-0068.1

Figure 8 shows the error correlations of surface meridional wind, surface air temperature, and SST in the CPL and CTL experiments. Corresponding model outputs of surface winds and sea level pressure are displayed in Fig. 8f. Because the error correlations at the eastern side of the TC center (cross marks) were evaluated where the horizontal gradient of sea level pressure was relatively steep, the horizontal distribution of the error correlation of the horizontal wind shows a cyclonic flow-dependent pattern in the atmosphere (Fig. 8a). A similar flow-dependent pattern can also be seen for horizontal wind as well as air temperature in the CTL experiment (Figs. 8d and 8e). However, in the CPL experiment, the surface air temperature correlation is less dependent on the atmospheric part of the flow but more dependent on the oceanic part of the vertical mixing, different from that in the CTL experiment (Fig. 8e). This would be owing to the strong correlation between SST and near-surface air temperature in the atmospheric boundary layer, actually corresponding to the horizontal distribution of the error correlation of SST (Fig. 8c).

Fig. 8.
Fig. 8.

Horizontal distribution of the error correlation (contour, every 1 hPa) of the (a) meridional surface wind, (b) surface air temperature, and (c) SST from the domain-center location (denoted by the cross marks, 134.0°N, 30.0°E) at 0000 UTC 9 Aug 2014 estimated in the CPL experiment. (d),(e) As in (a),(b), but for the CTL experiment. Contour intervals are 0.1. (f) Simulated mean sea level pressure (contour, every 4 hPa) and horizontal wind (m s−1, barbs) in the 6-h model forecast initialized from the LETKF analysis in the CPL experiment at the corresponding time.

Citation: Monthly Weather Review 145, 2; 10.1175/MWR-D-16-0068.1

In the CPL experiment, the oceanic variables were affected by the atmospheric variables through air–sea interactions. In general, the time scale of the variability of the oceanic component is larger than that of the atmosphere. For example, a cold wake induced by a TC was recovered over an e-folding time of 5–30 days (Price et al. 2008; Dare and McBride 2011). In fact, the data assimilation interval for oceanic models tends to be larger than one day, which is somewhat longer than that for atmospheric models (e.g., Zhang et al. 2012). Therefore, the present results imply that a high correlation between the atmospheric and the oceanic variables can possibly result from a delay of the construction of the atmospheric flow-dependent pattern of the error covariance for near-surface atmospheric variables, such as surface air temperature.

In addition, the coupling procedure in CNHM can be considered to be another factor for constructing an error covariance structure less dependent on the atmospheric part of the flow but more dependent on the oceanic part of the vertical mixing. Under a constant SST in NHM, air temperature in the atmospheric boundary layer strongly depends on turbulent heat fluxes determined by surface winds and the air–sea temperature difference determined only by air temperature. The effect of SST variations calculated by CNHM on turbulent heat fluxes and the resultant atmospheric boundary layer temperature becomes large. In particular, the effect is considered to be dominant along the track corresponding to the area of strong surface winds within the inner core of a TC because the area of obviously decreasing SST lingers along the track. In fact, the formation of the error covariance structure is similar to that of decreasing SST. The lower dependency on the atmospheric part of the flow in the atmospheric boundary layer might have a significant effect when observation data near the surface are assimilated where the flow-dependent error correlation becomes critical, such as in TC cases.

Figure 9 shows the results of the verification for 6-h first-guess forecasts relative to radiosonde observations averaged over 38 samples from 0000 UTC 1 August to 1200 UTC 19 August 2014 every 12 h. The influence of SST perturbations on the forecast was relatively small, but the impacts were positive on meridional wind, air temperature, and relative humidity. For the horizontal wind, the inclusion of SST perturbations resulted in a positive bias at the upper levels, leading to a reduction in RMSEs (Fig. 9a). More obvious improvement can be seen for air temperature biases and RMSEs around the 200-hPa level, whereas the bias increased at the lower levels (Fig. 9b). This verification result of air temperature bias is mainly due to the difference in the TC intensity between the CTL and CPL experiments. In the CTL experiment, the TC intensity was overestimated when it was located near Japan, whereas the intensity was closer to the observation in the CPL experiment (Fig. 10a). Therefore, the temperature anomaly near the TC center became smaller in the CPL experiment than in the CTL experiment, leading to a relatively low temperature profile at all levels in Fig. 9b. The improvement in the RMSE of the temperature is clear around the 200-hPa level, corresponding to the height of the warm core. This would be a manifestation of the smaller bias near the 200-hPa level in the CPL experiment. As for the moisture field, a small but positive impact was observed at all levels (Fig. 9c).

Fig. 9.
Fig. 9.

Verification of 6-h forecasts relative to radiosonde observations for (a) the meridional wind component (m s−1), (b) air temperature (K), and (c) relative humidity (%), averaged over 38 samples from 0000 UTC 1 Aug to 1200 UTC 19 Aug 2014 every 12 h. Blue lines indicate the results from the CTL experiment, and red lines indicate the results from the CPL experiment. Solid lines indicate RMSE and dashed lines indicate the bias.

Citation: Monthly Weather Review 145, 2; 10.1175/MWR-D-16-0068.1

Fig. 10.
Fig. 10.

Time series of the (a) minimum sea level pressure (central pressure) in the CTLM (blue), CPLM (red), and CPLO (green) experiments along with the best track data (black), and (b) track errors (km) relative to the best track data in the CTLM, CPLM, and CPLO experiments.

Citation: Monthly Weather Review 145, 2; 10.1175/MWR-D-16-0068.1

By comparing the vertical profile of the RMSE of the ensemble-mean forecast (Fig. 9) and the ensemble spread (Fig. 6), the impact of considering SST uncertainties in the NHM-LETKF system can be estimated in terms of the spread–skill relationship. Mainly due to the lack of representing model errors in the LETKF, the ensembles were underdispersive, that is, the ensemble spreads were smaller than the averaged RMSEs of the ensemble-mean forecasts. For the meridional wind and relative humidity, the influence of the inclusion of CNHM in the NHM-LETKF system on the spread–skill relationship would not be apparent because the change in the ensemble spread or the RMSEs of ensemble-mean forecast was relatively small. However, the ensemble spread of the air temperature increased at lower levels by including SST perturbations, which showed a better correspondence with the vertical profile of the RMSE of air temperature at lower levels.

d. Forecast verifications

The accuracy of the analyses for each experiment was verified with forecasts by NHM for TC Halong through the experiments in the NHM-LETKF system. As listed in Table 1, three experiments (see section 2c) were carried out for each configuration associated with atmospheric and oceanic initial states. Figure 10a shows the time series of minimum sea level pressure every 6 h for 48-h forecasts with eight different initial times from 0000 UTC 7 August to 1800 UTC 8 August 2014. In the following, TC intensity is defined as minimum sea level pressure. In all experiments, TC intensity tended to be weaker at each initial time than it would be in the best track analysis. This is probably because the horizontal resolution (15 km) was not sufficient in the data assimilation cycle for resolving the inner core of a TC, even though a horizontal resolution of 5 km was used for subsequent single deterministic NHM forecasts.

All deterministic results calculated by NHM in the CTLM experiments show that the TC intensified at the beginning of the forecast period, even though the best track analysis indicates the decaying phase. A similar spinup was also seen in all of the CPLM experiments. However, the impact of the spinup on TC intensity was slightly reduced for the deterministic results calculated by NHM in the CPLM experiments as compared to those in the CTLM experiments. This indicates that the TC intensity change in the CPLM experiments better corresponds to the best track analysis than in the CTLM experiments. In the CPLM experiments, the initial atmospheric states, particularly the air temperature and humidity fields at lower levels, were directly affected by air–sea interactions in the data assimilation cycle with CNHM. The cooler SST around the cold wake induced by Halong in the CPL experiment would alleviate the spinup in the early phase of the forecast. The alleviation of the spinup resulting from the use of CNHM is also obvious for all deterministic results calculated by NHM in the CPLO experiments because both atmospheric variables and SST were derived from the data assimilation cycle with CNHM. Although the simulated minimum sea level pressure in the CPLO experiments was higher than the best track analysis and, therefore, its mean absolute error relative to the best track data would be the worst, the tendency of the decaying phase of Halong was well captured. The higher horizontal resolution of the data assimilation system would help improve the mean absolute error of the simulated minimum sea level pressure in CPLO. Figure 10b depicts the averaged TC track errors relative to the best track observations. The CPLM experiments showed a slightly larger error than the CTLM experiments with short lead times. However, the track errors later improved. Although the CPLO forecasts exhibit a similar trend up to 18 h, subsequent forecasts had somewhat increased track errors compared to those in the CTLM experiments.

After the passage of TC Halong, heavy rainfall events continued intermittently throughout Japan because a stationary front extended across the entire country (not shown). Verification was conducted for precipitation forecasts to ascertain the overall impacts of the data assimilation cycle with CNHM. Here, the threat score (TS) and bias score (BS) were measured for 3-h accumulated rainfall amounts interpolated horizontally into 10-km grid boxes from the 5-km NHM outputs and the JMA’s radar/rain gauge–analyzed precipitation data (Nagata 2011). The verification scores are defined as
e2
e3
where F is the number of points when forecast precipitation exceeds a threshold, O is the number of points when observed precipitation exceeds a threshold, and H is the number of points when precipitation was correctly predicted.

Figure 11 shows the BS and TS over the land area of Japan for each 12- and 24-h forecast averaged over 32 samples from 0000 UTC 12 August to 1800 UTC 19 August 2014 at 6-h intervals. For the 12-h forecasts, no notable differences were found for the TS in the CPLM and CPLO experiments (Fig. 11a). However, the positive bias in the CTLM experiments was reduced for strong rainfall events compared to the CPLM and CPLO experiments (Fig. 11b). This is probably due to the fact that a more stable boundary layer caused by decreases in air–sea turbulent heat fluxes over the cold wake resulted in the improvement of the overestimation of precipitation over land areas through modification of moisture transport by southerly winds in the CPLM and CPLO experiments. For the 24-h forecasts, the advantage became small for the BS, but the verification result for the TS was substantially superior to that in the CTLM experiments. This result suggests that the improvement of the bias score from the 12- to 24-h forecasts would have greater influence on the improvement of the threat score in the 24-h forecast in CPLM and CPLO, compared with that in the CTLM experiments.

Fig. 11.
Fig. 11.

(a),(c) Threat scores and (b),(d) bias scores for 3-h accumulated precipitation amount averaged over 32 initial times from 0000 UTC 12 Aug to 1800 UTC 19 Aug for the 12- and 24-h deterministic NHM model forecasts in the CTLM (blue), CPLM (red), and CPLO (green) experiments. The scores were averaged over the land area of Japan.

Citation: Monthly Weather Review 145, 2; 10.1175/MWR-D-16-0068.1

The verification scores were also calculated for the grid points over the ocean (Fig. 12). Because observation data were scarce over the ocean compared to over the land area, mainly due to the lack of satellite observations in the current configuration, the TS was generally degraded compared to those in Fig. 11. As seen in Fig. 5c, SSTs in the CPLO experiments were cooler than those used in the CTLM and CPLM experiments during the verification period, which would lead to a negative bias of precipitation in the CPLO experiments (Figs. 12b and 12d). Despite the negative bias in the CPLO experiments, the TS was slightly superior to (Fig. 12a) or comparable with the other experiments (Fig. 12b). The verification results showed that the data assimilation cycle with CNHM contributed to the amelioration of precipitation scores over the land area, but the impact was relatively limited over the ocean, probably due to the negative bias of SST in the CPLO experiments.

Fig. 12.
Fig. 12.

(a),(c) Threat scores and (b),(d) bias scores for 3-h accumulated precipitation amount averaged over 32 initial times from 0000 UTC 12 Aug to 1800 UTC 19 Aug for the 12- and 24-h deterministic NHM model forecasts in the CTLM (blue), CPLM (red), and CPLO (green) experiments. The scores were averaged over the ocean.

Citation: Monthly Weather Review 145, 2; 10.1175/MWR-D-16-0068.1

4. Summary and discussion

The influence of SST estimation and its uncertainties on ensemble-based data assimilation was assessed by executing the NHM-LETKF system with an atmosphere–ocean coupled model. A simple configuration was applied in which only the atmospheric variables were updated in data assimilation; this was a preliminary test for the more complex coupled data assimilation involving both atmospheric and oceanic control variables. The brute-force application of the coupled model with the LETKF system for an experimental period more than one month caused a negative bias in SST. Tuning the bulk Richardson number in a vertically one-dimensional upper ocean model notably improved the SST bias, resulting in a better correspondence with the satellite-based OISST product. Inclusions of physically based SST variability in the system led to an increase in the ensemble spreads for air temperature and humidity at lower model levels because of positive correlations between SST and these atmospheric variables. It was also found that the horizontal distribution of error covariance for near-surface atmospheric variables became less dependent on the atmospheric part of the flow but more dependent on the oceanic part of the vertical mixing when the atmosphere–ocean coupled model was used. This was possibly due to the difference in time scales of the variability of the atmospheric and oceanic components as well as a relatively strong atmospheric boundary layer response to TC-induced sea surface cooling lingering along the track. The results indicate that the application of EnKF with the atmosphere–ocean coupled model could affect the assimilation of observations near the surface in severe weather events where the flow-dependent and in part mixing-dependent error covariance becomes essential in EnKF. Despite the limitation of the ocean model leading to the degradation of the SST estimate over the Kuroshio, statistical verifications based on radiosonde observations showed that the utilization of EnKF with the atmosphere–ocean coupled model contributed to the improvement of the errors as a sophisticated data assimilation system.

The system was also validated through verification of the extended deterministic model forecasts. For a TC event (TC Halong in 2014), the atmospheric fields derived from the EnKF system with the atmosphere–ocean coupled model contributed to alleviating the overintensification of Halong at the beginning of the forecast period. Simultaneous application of the analyzed atmosphere and SST fields as the initial conditions in the experiment by the NHM-LETKF system with the atmosphere–ocean coupled model helped to successfully simulate the decaying phase of Halong analyzed in the Regional Specialized Meteorological Center Tokyo best track dataset, but the TC intensity was weaker than that of the best track analysis, possibly because of insufficient horizontal resolution of the data assimilation system. Verification scores for precipitation amounts were also improved by using the NHM-LETKF system with the atmosphere–ocean coupled model, but the effect was relatively limited.

The bulk Richardson number in the ocean model was an important parameter for controlling the SST bias with relatively longer lead times. Although an explicit correction of the SST bias through the data assimilation of oceanic observations would contribute to reducing the bias in the analyses, the correction of the model bias itself would be essential for data assimilation. In the framework of EnKF, it has been noted that optimal model parameters are estimated simultaneously with model variables by taking the error covariance between model parameters and observations into consideration (Aksoy et al. 2006; Anderson 2001). Recent studies have demonstrated that the ensemble-based parameter estimation with an atmosphere–ocean coupled model is effective (Zhang et al. 2012; Wu et al. 2012; Liu et al. 2014). Further studies on the sensitivities of the parameters in the atmosphere–ocean coupled model are expected to contribute to the improvement of the system and forecasts in the future.

A limitation of the present study is that the oceanic variables were not treated as control variables. Therefore, the impact of oceanic observations could not be utilized for updating the atmospheric states through a data assimilation cycle in this study. Recently, the number of oceanic observations has been increasing, including special observation campaigns such as the Impacts of Typhoons on the Ocean in the Pacific program in 2010 (D’Asaro et al. 2011; Pun et al. 2011). Even though the observation network over the ocean has been limited in general, the utilization of oceanic observations can be expected to contribute to improving the accuracy of atmospheric data assimilation, particularly over the ocean. In that sense, a strong coupling that assimilates oceanic variables is required to update atmospheric states when the analysis increments for atmospheric and oceanic fields are simultaneously calculated. Although the implementation of strong coupling in the EnKF with an atmosphere–ocean coupled model is easy in principle, it is still under debate. Further studies will be required to utilize oceanic observations to improve numerical weather predictions.

Acknowledgments

We thank the members of the Forecast Research Department of the Meteorological Research Institute for fruitful discussions. This work was supported by a Japan Society for the Promotion of Science Grant-in-Aid for Young Scientists (B) Grant 26800247 and for Scientific Research (C) Grant 15K05292, as well as by the Strategic Programs for Innovative Research (SPIRE). Some of the results were obtained by using the K computer at the RIKEN Advanced Institute for Computational Science (Proposal hp140220).

REFERENCES

  • Aksoy, A., F. Zhang, and J. W. Nielsen-Gammon, 2006: Ensemble-based simultaneous state and parameter estimation in a two-dimensional sea-breeze model. Mon. Wea. Rev., 134, 29512970, doi:10.1175/MWR3224.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Anderson, J. L., 2001: An ensemble adjustment Kalman filter for data assimilation. Mon. Wea. Rev., 129, 28842903, doi:10.1175/1520-0493(2001)129<2884:AEAKFF>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Barker, D. M., 2005: Southern high-latitude ensemble data assimilation in the Antarctic Mesoscale Prediction System. Mon. Wea. Rev., 133, 34313449, doi:10.1175/MWR3042.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bonavita, M., L. Torrisi, and F. Marcucci, 2008: The ensemble Kalman filter in an operational regional NWP system: Preliminary results with real observations. Quart. J. Roy. Meteor. Soc., 134, 17331744, doi:10.1002/qj.313.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bowler, N. E., and K. R. Mylne, 2009: Ensemble transform Kalman filter perturbation for a regional ensemble prediction system. Quart. J. Roy. Meteor. Soc., 135, 757766, doi:10.1002/qj.404.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Chen, S. S., W. Zhao, M. A. Donelan, and H. L. Tolman, 2013: Directional wind–wave coupling in fully coupled atmosphere–wave–ocean models: Results from CBLAST-Hurricane. J. Atmos. Sci., 70, 31983215, doi:10.1175/JAS-D-12-0157.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Courtier, P., J.-N. Thepaut, and A. Hollingsworth, 1994: A strategy for operational implementation of 4D-Var using an incremental approach. Quart. J. Roy. Meteor. Soc., 120, 13671387, doi:10.1002/qj.49712051912.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • D’Asaro, E., and Coauthors, 2011: Typhoon–ocean interaction in the western North Pacific: Part 1. Oceanography, 24, 2431, doi:10.5670/oceanog.2011.91.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dare, R. A., and J. L. McBride, 2011: Sea surface temperature response to tropical cyclones. Mon. Wea. Rev., 139, 37983838, doi:10.1175/MWR-D-10-05019.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dowell, D. C., F. Zhang, L. J. Wicker, C. Snyder, and N. A. Crook, 2004: Wind and temperature retrievals in the 17 May 1981 Arcadia, Oklahoma, supercell: Ensemble Kalman filter experiments. Mon. Wea. Rev., 132, 19822005, doi:10.1175/1520-0493(2004)132<1982:WATRIT>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Evensen, G., 1994: Sequential data assimilation with a nonlinear quasi-geostrophic model using Monte Carlo methods to forecast error statistics. J. Geophys. Res., 99, 10 14310 162, doi:10.1029/94JC00572.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hamill, T. M., J. S. Whitaker, M. Fiorino, and S. G. Benjamin, 2011: Global ensemble predictions of 2009’s tropical cyclones initialized with an ensemble Kalman filter. Mon. Wea. Rev., 139, 668688, doi:10.1175/2010MWR3456.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hou, D., E. Kalnay, and K. K. Droegemeier, 2001: Objective verification of the SAMEX-98 ensemble forecasts. Mon. Wea. Rev., 129, 7391, doi:10.1175/1520-0493(2001)129<0073:OVOTSE>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hunt, B. R., E. J. Kostelich, and I. Syzunogh, 2007: Efficient data assimilation for spatiotemporal chaos: A local ensemble transform Kalman filter. Physica D, 230, 112126, doi:10.1016/j.physd.2006.11.008.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ikawa, M., and K. Saito, 1991: Description of a nonhydrostatic model developed at the Forecast Research Department of the MRI. MRI Tech. Rep. 28, 238 pp.

  • Ito, K., T. Kuroda, K. Saito, and A. Wada, 2015: Forecasting a large number of tropical cyclone intensities around Japan using a high-resolution atmosphere–ocean coupled model. Wea. Forecasting, 30, 793808, doi:10.1175/WAF-D-14-00034.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Japan Meteorological Agency, 2013: Outline of the operational numerical weather prediction at the Japan Meteorological Agency. Appendix to WMO technical progress report on the global data-processing and forecasting system and numerical weather prediction. Japan Meteorological Agency, 187 pp. [Available online at http://www.jma.go.jp/jma/jma-eng/jma-center/nwp/outline2013-nwp/index.htm.]

  • Kawai, Y., and A. Wada, 2007: Diurnal sea surface temperature variation and its impact on the atmosphere and ocean: A review. J. Oceanogr., 63, 721744, doi:10.1007/s10872-007-0063-0.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kunii, M., 2014a: Mesoscale data assimilation for a local severe rainfall event with the NHM–LETKF system. Wea. Forecasting, 29, 10931105, doi:10.1175/WAF-D-13-00032.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kunii, M., 2014b: The 1000-member ensemble Kalman filtering with the JMA nonhydrostatic mesoscale model on the K computer. J. Meteor. Soc. Japan, 92, 623633, doi:10.2151/jmsj.2014-607.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kunii, M., and T. Miyoshi, 2012: Including uncertainties of sea surface temperature in an ensemble Kalman filter: A case study of Typhoon Sinlaku (2008). Wea. Forecasting, 27, 15861597, doi:10.1175/WAF-D-11-00136.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kurihara, Y., T. Sakurai, and T. Kuragano, 2006: Global daily sea surface temperature analysis using data from satellite microwave radiometer, satellite infrared radiometer and in-situ observations (in Japanese). Wea. Service Bull., 73, 118.

    • Search Google Scholar
    • Export Citation
  • Lewis, J. M., and J. C. Derber, 1985: The use of adjoint equations to solve a variational adjustment problem with advective constraints. Tellus, 37A, 309322, doi:10.1111/j.1600-0870.1985.tb00430.x.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Liu, Y., Z. Liu, S. Zhang, R. Jacob, F. Lu, X. Rong, and S. Wu, 2014: Ensemble-based parameter estimation in a coupled general circulation model. J. Climate, 27, 71517162, doi:10.1175/JCLI-D-13-00406.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • McLay, J. G., M. K. Flatau, C. A. Reynolds, J. Cummings, T. Hogan, and P. J. Flatau, 2012: Inclusion of sea-surface temperature variation in the U.S. Navy ensemble-transform global ensemble prediction system. J. Geophys. Res., 117, D19120, doi:10.1029/2011JD016937.

    • Search Google Scholar
    • Export Citation
  • Miyoshi, T., and M. Kunii, 2012: The local ensemble transform Kalman filter with the Weather Research and Forecasting model: Experiments with real observations. Pure Appl. Geophys., 169, 321333, doi:10.1007/s00024-011-0373-4.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Nagata, K., 2011: Quantitative precipitation estimation and quantitative precipitation forecasting by the Japan meteorological agency. RSMC Tokyo–Typhoon Cent. Tech. Rev.,13, 37–50. [Available online at http://www.jma.go.jp/jma/jma-eng/jma-center/rsmc-hp-pub-eg/techrev/text13-2.pdf.]

  • Nakanishi, M., and H. Niino, 2004: An improved Mellor-Yamada level-3 model with condensation physics: Its design and verification. Bound.-Layer Meteor., 112, 131, doi:10.1023/B:BOUN.0000020164.04146.98.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Nakanishi, M., and H. Niino, 2006: An improved Mellor-Yamada level-3 model: Its numerical stability and application to a regional prediction of advection fog. Bound.-Layer Meteor., 119, 397407, doi:10.1007/s10546-005-9030-8.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ohmori, S., and Y. Yamada, 2006: Development of cumulus parameterization scheme in the nonhydrostatic mesoscale model at the Japan Meteorological Agency. CAS/JSC WGNE Res. Activ. Atmos. Oceanic Modell., 35, 4.214.22.

    • Search Google Scholar
    • Export Citation
  • Ouaraini, R. E., L. Berre, C. Fischer, and E. H. Sayouty, 2015: Sensitivity of regional ensemble data assimilation spread to perturbations of lateral boundary conditions. Tellus, 67A, 28502, doi:10.3402/tellusa.v67.28502.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Pendergrass, A. G., G. J. Hakim, D. S. Battisti, and G. Roe, 2012: Coupled air–mixed layer temperature predictability for climate reconstruction. J. Climate, 25, 459472, doi:10.1175/2011JCLI4094.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Price, J. F., R. A. Weller, and R. Pinkel, 1986: Diurnal cycling: Observations and models of the upper ocean response to diurnal heating, cooling, and wind mixing. J. Geophys. Res., 91, 84118427, doi:10.1029/JC091iC07p08411.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Price, J. F., T. Sanford, and G. Forristall, 1994: Forced stage response to a moving hurricane. J. Phys. Oceanogr., 24, 233260, doi:10.1175/1520-0485(1994)024<0233:FSRTAM>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Price, J. F., J. Morzel, and P. P. Niiler, 2008: Warming of SST in the cool wake of a moving hurricane. J. Geophys. Res., 113, C07010, doi:10.1029/2007JC004393.

    • Search Google Scholar
    • Export Citation
  • Pun, I.-F., Y.-T. Chang, I.-I. Lin, T. Y. Tang, and R.-C. Lien, 2011: Typhoon-ocean interaction in the western North Pacific: Part 2. Oceanography, 24, 3241, doi:10.5670/oceanog.2011.92.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Saito, K., 2012: The Japan Meteorological Agency nonhydrostatic model and its application to operation and research. Atmospheric Model Applications, I. Yucel, Ed., InTech, 85–110, doi:10.5772/35368.

    • Crossref
    • Export Citation
  • Saito, K., and Coauthors, 2006: The operational JMA nonhydrostatic mesoscale model. Mon. Wea. Rev., 134, 12661298, doi:10.1175/MWR3120.1.

  • Saito, K., J. Ishida, K. Aranami, T. Hara, T. Segawa, M. Narita, and Y. Honda, 2007: Nonhydrostatic atmospheric models and operational development at JMA. J. Meteor. Soc. Japan, 85B, 271304, doi:10.2151/jmsj.85B.271.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Saito, K., H. Seko, M. Kunii, and T. Miyoshi, 2012: Effect of lateral boundary perturbations on the breeding method and the local ensemble transform Kalman filter for mesoscale ensemble prediction. Tellus, 64A, 11594, doi:10.3402/tellusa.v64i0.11594.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Schwartz, C. S., G. S. Romine, M. L. Weisman, R. A. Sobash, K. R. Fossell, K. W. Manning, and S. B. Trier, 2015: A real-time convection-allowing ensemble prediction system initialized by mesoscale ensemble Kalman filter analyses. Wea. Forecasting, 30, 11581181, doi:10.1175/WAF-D-15-0013.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Sekiyama, T. T., M. Kunii, M. Kajino, and T. Shimbori, 2015: Horizontal resolution dependence of atmospheric simulations of the Fukushima nuclear accident using 15-km, 3-km, and 500-m grid models. J. Meteor. Soc. Japan, 93, 4964, doi:10.2151/jmsj.2015-002.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Snyder, C., and F. Zhang, 2003: Assimilation of simulated Doppler radar observations with an ensemble Kalman filter. Mon. Wea. Rev., 131, 16631677, doi:10.1175//2555.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Tardif, R., G. J. Hakim, and C. Snyder, 2014: Coupled atmosphere-ocean data assimilation experiments with a low-order climate model. Climate Dyn., 43, 16311643, doi:10.1007/s00382-013-1989-0.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Torn, R. D., G. J. Hakim, and C. Snyder, 2006: Boundary conditions for limited-area ensemble Kalman filters. Mon. Wea. Rev., 134, 24902502, doi:10.1175/MWR3187.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wada, A., and M. Kunii, 2014: Introduction of an atmosphere-wave-ocean coupled model into the NHM-LETKF. CAS/JSC WGNE Res. Activ. Atmos. Oceanic Modell., 44, 9.039.04.

    • Search Google Scholar
    • Export Citation
  • Wada, A., T. Uehara, and S. Ishizaki, 2014: Typhoon-induced sea surface cooling during the 2011 and 2012 typhoon seasons: Observational evidence and numerical investigations of the sea surface cooling effect using typhoon simulations. Prog. Earth Planet. Sci., 1, 11, doi:10.1186/2197-4284-1-11.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Whitaker, J. S., and T. M. Hamill, 2012: Evaluating methods to account for system errors in ensemble data assimilation. Mon. Wea. Rev., 140, 30783089, doi:10.1175/MWR-D-11-00276.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wu, X., S. Zhang, Z. Liu, A. Rosati, T. Delworth, and Y. Liu, 2012: Impact of geographic-dependent parameter optimization on climate estimation and prediction: Simulation with an intermediate coupled model. Mon. Wea. Rev., 140, 39563971, doi:10.1175/MWR-D-11-00298.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zhang, S., M. J. Harrison, A. Rosati, and A. Wittenberg, 2007: System design and evaluation of coupled ensemble data assimilation for global oceanic climate studies. Mon. Wea. Rev., 135, 35413564, doi:10.1175/MWR3466.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zhang, S., Z. Liu, A. Rosati, and T. Delworth, 2012: A study of enhancive parameter correction with coupled data assimilation for climate estimation and prediction using a simple coupled model. Tellus, 64A, 10963, doi:10.3402/tellusa.v64i0.10963.

    • Crossref
    • Search Google Scholar
    • Export Citation
1

In Ito et al. (2015), the coupled model was referred to as “CMSM.”

Save
  • Aksoy, A., F. Zhang, and J. W. Nielsen-Gammon, 2006: Ensemble-based simultaneous state and parameter estimation in a two-dimensional sea-breeze model. Mon. Wea. Rev., 134, 29512970, doi:10.1175/MWR3224.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Anderson, J. L., 2001: An ensemble adjustment Kalman filter for data assimilation. Mon. Wea. Rev., 129, 28842903, doi:10.1175/1520-0493(2001)129<2884:AEAKFF>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Barker, D. M., 2005: Southern high-latitude ensemble data assimilation in the Antarctic Mesoscale Prediction System. Mon. Wea. Rev., 133, 34313449, doi:10.1175/MWR3042.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bonavita, M., L. Torrisi, and F. Marcucci, 2008: The ensemble Kalman filter in an operational regional NWP system: Preliminary results with real observations. Quart. J. Roy. Meteor. Soc., 134, 17331744, doi:10.1002/qj.313.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bowler, N. E., and K. R. Mylne, 2009: Ensemble transform Kalman filter perturbation for a regional ensemble prediction system. Quart. J. Roy. Meteor. Soc., 135, 757766, doi:10.1002/qj.404.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Chen, S. S., W. Zhao, M. A. Donelan, and H. L. Tolman, 2013: Directional wind–wave coupling in fully coupled atmosphere–wave–ocean models: Results from CBLAST-Hurricane. J. Atmos. Sci., 70, 31983215, doi:10.1175/JAS-D-12-0157.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Courtier, P., J.-N. Thepaut, and A. Hollingsworth, 1994: A strategy for operational implementation of 4D-Var using an incremental approach. Quart. J. Roy. Meteor. Soc., 120, 13671387, doi:10.1002/qj.49712051912.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • D’Asaro, E., and Coauthors, 2011: Typhoon–ocean interaction in the western North Pacific: Part 1. Oceanography, 24, 2431, doi:10.5670/oceanog.2011.91.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dare, R. A., and J. L. McBride, 2011: Sea surface temperature response to tropical cyclones. Mon. Wea. Rev., 139, 37983838, doi:10.1175/MWR-D-10-05019.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dowell, D. C., F. Zhang, L. J. Wicker, C. Snyder, and N. A. Crook, 2004: Wind and temperature retrievals in the 17 May 1981 Arcadia, Oklahoma, supercell: Ensemble Kalman filter experiments. Mon. Wea. Rev., 132, 19822005, doi:10.1175/1520-0493(2004)132<1982:WATRIT>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Evensen, G., 1994: Sequential data assimilation with a nonlinear quasi-geostrophic model using Monte Carlo methods to forecast error statistics. J. Geophys. Res., 99, 10 14310 162, doi:10.1029/94JC00572.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hamill, T. M., J. S. Whitaker, M. Fiorino, and S. G. Benjamin, 2011: Global ensemble predictions of 2009’s tropical cyclones initialized with an ensemble Kalman filter. Mon. Wea. Rev., 139, 668688, doi:10.1175/2010MWR3456.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hou, D., E. Kalnay, and K. K. Droegemeier, 2001: Objective verification of the SAMEX-98 ensemble forecasts. Mon. Wea. Rev., 129, 7391, doi:10.1175/1520-0493(2001)129<0073:OVOTSE>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hunt, B. R., E. J. Kostelich, and I. Syzunogh, 2007: Efficient data assimilation for spatiotemporal chaos: A local ensemble transform Kalman filter. Physica D, 230, 112126, doi:10.1016/j.physd.2006.11.008.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ikawa, M., and K. Saito, 1991: Description of a nonhydrostatic model developed at the Forecast Research Department of the MRI. MRI Tech. Rep. 28, 238 pp.

  • Ito, K., T. Kuroda, K. Saito, and A. Wada, 2015: Forecasting a large number of tropical cyclone intensities around Japan using a high-resolution atmosphere–ocean coupled model. Wea. Forecasting, 30, 793808, doi:10.1175/WAF-D-14-00034.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Japan Meteorological Agency, 2013: Outline of the operational numerical weather prediction at the Japan Meteorological Agency. Appendix to WMO technical progress report on the global data-processing and forecasting system and numerical weather prediction. Japan Meteorological Agency, 187 pp. [Available online at http://www.jma.go.jp/jma/jma-eng/jma-center/nwp/outline2013-nwp/index.htm.]

  • Kawai, Y., and A. Wada, 2007: Diurnal sea surface temperature variation and its impact on the atmosphere and ocean: A review. J. Oceanogr., 63, 721744, doi:10.1007/s10872-007-0063-0.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kunii, M., 2014a: Mesoscale data assimilation for a local severe rainfall event with the NHM–LETKF system. Wea. Forecasting, 29, 10931105, doi:10.1175/WAF-D-13-00032.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kunii, M., 2014b: The 1000-member ensemble Kalman filtering with the JMA nonhydrostatic mesoscale model on the K computer. J. Meteor. Soc. Japan, 92, 623633, doi:10.2151/jmsj.2014-607.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kunii, M., and T. Miyoshi, 2012: Including uncertainties of sea surface temperature in an ensemble Kalman filter: A case study of Typhoon Sinlaku (2008). Wea. Forecasting, 27, 15861597, doi:10.1175/WAF-D-11-00136.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kurihara, Y., T. Sakurai, and T. Kuragano, 2006: Global daily sea surface temperature analysis using data from satellite microwave radiometer, satellite infrared radiometer and in-situ observations (in Japanese). Wea. Service Bull., 73, 118.

    • Search Google Scholar
    • Export Citation
  • Lewis, J. M., and J. C. Derber, 1985: The use of adjoint equations to solve a variational adjustment problem with advective constraints. Tellus, 37A, 309322, doi:10.1111/j.1600-0870.1985.tb00430.x.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Liu, Y., Z. Liu, S. Zhang, R. Jacob, F. Lu, X. Rong, and S. Wu, 2014: Ensemble-based parameter estimation in a coupled general circulation model. J. Climate, 27, 71517162, doi:10.1175/JCLI-D-13-00406.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • McLay, J. G., M. K. Flatau, C. A. Reynolds, J. Cummings, T. Hogan, and P. J. Flatau, 2012: Inclusion of sea-surface temperature variation in the U.S. Navy ensemble-transform global ensemble prediction system. J. Geophys. Res., 117, D19120, doi:10.1029/2011JD016937.

    • Search Google Scholar
    • Export Citation
  • Miyoshi, T., and M. Kunii, 2012: The local ensemble transform Kalman filter with the Weather Research and Forecasting model: Experiments with real observations. Pure Appl. Geophys., 169, 321333, doi:10.1007/s00024-011-0373-4.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Nagata, K., 2011: Quantitative precipitation estimation and quantitative precipitation forecasting by the Japan meteorological agency. RSMC Tokyo–Typhoon Cent. Tech. Rev.,13, 37–50. [Available online at http://www.jma.go.jp/jma/jma-eng/jma-center/rsmc-hp-pub-eg/techrev/text13-2.pdf.]

  • Nakanishi, M., and H. Niino, 2004: An improved Mellor-Yamada level-3 model with condensation physics: Its design and verification. Bound.-Layer Meteor., 112, 131, doi:10.1023/B:BOUN.0000020164.04146.98.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Nakanishi, M., and H. Niino, 2006: An improved Mellor-Yamada level-3 model: Its numerical stability and application to a regional prediction of advection fog. Bound.-Layer Meteor., 119, 397407, doi:10.1007/s10546-005-9030-8.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ohmori, S., and Y. Yamada, 2006: Development of cumulus parameterization scheme in the nonhydrostatic mesoscale model at the Japan Meteorological Agency. CAS/JSC WGNE Res. Activ. Atmos. Oceanic Modell., 35, 4.214.22.

    • Search Google Scholar
    • Export Citation
  • Ouaraini, R. E., L. Berre, C. Fischer, and E. H. Sayouty, 2015: Sensitivity of regional ensemble data assimilation spread to perturbations of lateral boundary conditions. Tellus, 67A, 28502, doi:10.3402/tellusa.v67.28502.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Pendergrass, A. G., G. J. Hakim, D. S. Battisti, and G. Roe, 2012: Coupled air–mixed layer temperature predictability for climate reconstruction. J. Climate, 25, 459472, doi:10.1175/2011JCLI4094.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Price, J. F., R. A. Weller, and R. Pinkel, 1986: Diurnal cycling: Observations and models of the upper ocean response to diurnal heating, cooling, and wind mixing. J. Geophys. Res., 91, 84118427, doi:10.1029/JC091iC07p08411.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Price, J. F., T. Sanford, and G. Forristall, 1994: Forced stage response to a moving hurricane. J. Phys. Oceanogr., 24, 233260, doi:10.1175/1520-0485(1994)024<0233:FSRTAM>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Price, J. F., J. Morzel, and P. P. Niiler, 2008: Warming of SST in the cool wake of a moving hurricane. J. Geophys. Res., 113, C07010, doi:10.1029/2007JC004393.

    • Search Google Scholar
    • Export Citation
  • Pun, I.-F., Y.-T. Chang, I.-I. Lin, T. Y. Tang, and R.-C. Lien, 2011: Typhoon-ocean interaction in the western North Pacific: Part 2. Oceanography, 24, 3241, doi:10.5670/oceanog.2011.92.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Saito, K., 2012: The Japan Meteorological Agency nonhydrostatic model and its application to operation and research. Atmospheric Model Applications, I. Yucel, Ed., InTech, 85–110, doi:10.5772/35368.

    • Crossref
    • Export Citation
  • Saito, K., and Coauthors, 2006: The operational JMA nonhydrostatic mesoscale model. Mon. Wea. Rev., 134, 12661298, doi:10.1175/MWR3120.1.

  • Saito, K., J. Ishida, K. Aranami, T. Hara, T. Segawa, M. Narita, and Y. Honda, 2007: Nonhydrostatic atmospheric models and operational development at JMA. J. Meteor. Soc. Japan, 85B, 271304, doi:10.2151/jmsj.85B.271.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Saito, K., H. Seko, M. Kunii, and T. Miyoshi, 2012: Effect of lateral boundary perturbations on the breeding method and the local ensemble transform Kalman filter for mesoscale ensemble prediction. Tellus, 64A, 11594, doi:10.3402/tellusa.v64i0.11594.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Schwartz, C. S., G. S. Romine, M. L. Weisman, R. A. Sobash, K. R. Fossell, K. W. Manning, and S. B. Trier, 2015: A real-time convection-allowing ensemble prediction system initialized by mesoscale ensemble Kalman filter analyses. Wea. Forecasting, 30, 11581181, doi:10.1175/WAF-D-15-0013.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Sekiyama, T. T., M. Kunii, M. Kajino, and T. Shimbori, 2015: Horizontal resolution dependence of atmospheric simulations of the Fukushima nuclear accident using 15-km, 3-km, and 500-m grid models. J. Meteor. Soc. Japan, 93, 4964, doi:10.2151/jmsj.2015-002.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Snyder, C., and F. Zhang, 2003: Assimilation of simulated Doppler radar observations with an ensemble Kalman filter. Mon. Wea. Rev., 131, 16631677, doi:10.1175//2555.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Tardif, R., G. J. Hakim, and C. Snyder, 2014: Coupled atmosphere-ocean data assimilation experiments with a low-order climate model. Climate Dyn., 43, 16311643, doi:10.1007/s00382-013-1989-0.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Torn, R. D., G. J. Hakim, and C. Snyder, 2006: Boundary conditions for limited-area ensemble Kalman filters. Mon. Wea. Rev., 134, 24902502, doi:10.1175/MWR3187.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wada, A., and M. Kunii, 2014: Introduction of an atmosphere-wave-ocean coupled model into the NHM-LETKF. CAS/JSC WGNE Res. Activ. Atmos. Oceanic Modell., 44, 9.039.04.

    • Search Google Scholar
    • Export Citation
  • Wada, A., T. Uehara, and S. Ishizaki, 2014: Typhoon-induced sea surface cooling during the 2011 and 2012 typhoon seasons: Observational evidence and numerical investigations of the sea surface cooling effect using typhoon simulations. Prog. Earth Planet. Sci., 1, 11, doi:10.1186/2197-4284-1-11.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Whitaker, J. S., and T. M. Hamill, 2012: Evaluating methods to account for system errors in ensemble data assimilation. Mon. Wea. Rev., 140, 30783089, doi:10.1175/MWR-D-11-00276.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wu, X., S. Zhang, Z. Liu, A. Rosati, T. Delworth, and Y. Liu, 2012: Impact of geographic-dependent parameter optimization on climate estimation and prediction: Simulation with an intermediate coupled model. Mon. Wea. Rev., 140, 39563971, doi:10.1175/MWR-D-11-00298.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zhang, S., M. J. Harrison, A. Rosati, and A. Wittenberg, 2007: System design and evaluation of coupled ensemble data assimilation for global oceanic climate studies. Mon. Wea. Rev., 135, 35413564, doi:10.1175/MWR3466.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zhang, S., Z. Liu, A. Rosati, and T. Delworth, 2012: A study of enhancive parameter correction with coupled data assimilation for climate estimation and prediction using a simple coupled model. Tellus, 64A, 10963, doi:10.3402/tellusa.v64i0.10963.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fig. 1.

    Model domain for the NHM-LETKF analysis and the subsequent model forecast. Colors indicate land elevation in meters.

  • Fig. 2.

    Time series of SST averaged over the model domain obtained from daily mean SST (MGDSST), ensemble mean SST in the CPL experiment (CPL), daily mean SST (OISST) based on satellite observations (OBS), and the ensemble spread of SST in the CPL experiment (Spread). The critical bulk Richardson numbers were (a) 0.6 and (b) 0.25.

  • Fig. 3.

    Vertical cross sections of the differences in ocean (a) temperature (K), (b) salinity, and (c) zonal current velocity (m s−1) between the CPL experiments with critical bulk Richardson numbers of 0.25 and 0.6 as a function of depth (m) and longitude at 30°N at 0000 UTC 10 Aug 2014. The contour interval is 0.2.

  • Fig. 4.

    Horizontal distributions of (a) daily mean SST obtained from the MGDSST, (b) the ensemble mean SST in the CPL experiment, and (c) daily mean SST obtained from the OISST on 10 Aug 2014. Shades and contours indicate the value of SST. The contour interval is 1°C.

  • Fig. 5.

    Horizontal distributions of RMSEs of (a) daily mean SST obtained from the MGDSST and (b) the ensemble mean SST in the CPL experiment relative to OISST, averaged over 19 days from 0000 UTC 1 Aug to 1800 UTC 19 Aug 2014. (c) Time series of bias and RMSE of the daily mean SST obtained from the MGDSST (black) and the ensemble mean SST in the CPL experiment (red) relative to OISST, averaged over the region 25°–40°N, 125°–145°E.

  • Fig. 6.

    Vertical profiles of the ensemble spread for (a) meridional wind component (m s−1), (b) temperature (K), and (c) water vapor mixing ratio (g kg−1), averaged over the model domain for 76 samples from 0000 UTC 1 Aug to 1800 UTC 19 Aug 2014 every 6 h. Blue dashed lines indicate the results from the CTL experiment. Red solid lines indicate the results from the CPL experiment.

  • Fig. 7.

    Vertical cross sections of the error correlation between the SST from the domain-center location (denoted by the cross marks, 32.5°N, 140.0°E) and (a) meridional wind, (b) air temperature, (c) water vapor mixing ratio, and (d) vertical wind. The correlation was an average of 76 samples from 0000 UTC 1 Aug to 1800 UTC 19 Aug 2014 every 6 h. Contour intervals are 0.1.

  • Fig. 8.

    Horizontal distribution of the error correlation (contour, every 1 hPa) of the (a) meridional surface wind, (b) surface air temperature, and (c) SST from the domain-center location (denoted by the cross marks, 134.0°N, 30.0°E) at 0000 UTC 9 Aug 2014 estimated in the CPL experiment. (d),(e) As in (a),(b), but for the CTL experiment. Contour intervals are 0.1. (f) Simulated mean sea level pressure (contour, every 4 hPa) and horizontal wind (m s−1, barbs) in the 6-h model forecast initialized from the LETKF analysis in the CPL experiment at the corresponding time.

  • Fig. 9.

    Verification of 6-h forecasts relative to radiosonde observations for (a) the meridional wind component (m s−1), (b) air temperature (K), and (c) relative humidity (%), averaged over 38 samples from 0000 UTC 1 Aug to 1200 UTC 19 Aug 2014 every 12 h. Blue lines indicate the results from the CTL experiment, and red lines indicate the results from the CPL experiment. Solid lines indicate RMSE and dashed lines indicate the bias.

  • Fig. 10.

    Time series of the (a) minimum sea level pressure (central pressure) in the CTLM (blue), CPLM (red), and CPLO (green) experiments along with the best track data (black), and (b) track errors (km) relative to the best track data in the CTLM, CPLM, and CPLO experiments.

  • Fig. 11.

    (a),(c) Threat scores and (b),(d) bias scores for 3-h accumulated precipitation amount averaged over 32 initial times from 0000 UTC 12 Aug to 1800 UTC 19 Aug for the 12- and 24-h deterministic NHM model forecasts in the CTLM (blue), CPLM (red), and CPLO (green) experiments. The scores were averaged over the land area of Japan.

  • Fig. 12.

    (a),(c) Threat scores and (b),(d) bias scores for 3-h accumulated precipitation amount averaged over 32 initial times from 0000 UTC 12 Aug to 1800 UTC 19 Aug for the 12- and 24-h deterministic NHM model forecasts in the CTLM (blue), CPLM (red), and CPLO (green) experiments. The scores were averaged over the ocean.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 489 102 14
PDF Downloads 459 51 8