Title:

Kawasaki Disease and ENSO-driven wind circulation

Authors:

Joan Ballester (1), Jane C. Burns (2), Dan Cayan (3), Yosikazu Nakamura (4), Ritei Uehara (4), Xavier Rodó (1,5)

Affiliations:

(1) Institut Català de Ciències del Clima (IC3), Barcelona, Catalonia, Spain

(2) Department of Pediatrics, Rady Children’s Hospital San Diego and UCSD, La Jolla, CA

(3) Climate, Atmospheric Science, and Physical Oceanography, Scripps Institution of Oceanography, UCSD and Water Resources Discipline, US Geological Survey, La Jolla, CA

(4) Department of Public Health, Jichi Medical University, Tochigi, Japan

(5) Institució Catalana de Recerca i Estudis Avançats (ICREA), Barcelona, Catalonia, Spain

Corresponding author (present address):

Joan Ballester

California Institute of Technology (Caltech)

1200 E California Blvd, Pasadena, CA 91125, US

Mail Code: 131-24

Tel.: +1 626-395-8703

Email:

Key points:

Kawasaki Disease is the most common cause of acquired heart disease in children

Kawasaki Disease may be triggered by a windborne agent

El Niño-Southern Oscillation is associated with enhanced Kawasaki Disease

Index terms:

3309 Climatology

3339 Ocean/atmosphere interactions

3373 Tropical dynamics

4215 Climate and interannual variability

4522 ENSO

Keywords:

Kawasaki Disease

El Niño-Southern Oscillation

El Niño and La Niña

Tropospheric winds

Tropical-extratropical teleconnections

Surface ocean conditions

Supplementary Material

Tropospheric mechanism explaining the seasonality and epidemics of KD

Rodó et al. (2011) described the relationship between the nearly synchronized winter peak of KD in Japan and San Diego and the wintertime tropospheric “duct” sweeping from continental Asia to the western north Pacific (green box in Supplementary Figure 1). The close timing of the seasonal peak in both sites, as well as its synchrony with both strong low-level northwesterly winds in Japan and intense zonal winds throughout the subtropical north Pacific, suggested that the causal agent of KD is transported from continental Asia to Japan by low-level air currents, and across the north Pacific to the United States by strong zonal winds developing in the upper troposphere. This tropospheric winter window coincided with the growing phase of 3 dramatic nationwide epidemics in Japan (Supplementary Figure 2), by far the largest epidemic events ever recorded worldwide. Thus, the number of cases rapidly mounted in autumn all over the country when winds turned northwest in direction, and suddenly started to decrease in spring when winds again shifted and blew from the south.

KD time series in Japan and San Diego

The Japanese KD time series is derived from the biannual national surveillance of all hospitals in Japan with more than 100 beds. Data corresponds to 16 separate questionnaire surveys of Japanese hospitals for the period 1970–2010, providing the most comprehensive record of KD cases in the world. Patient date of hospitalization and prefecture of residence are recorded for each subject in the time series, which contains 271,415 cases until 2010 (141,677 cases during the period 1987-2006 considered here).

The time series for hospital admissions in San Diego was assembled from the database of the Kawasaki Disease Research Center at the University of California for the period 1994–2008. It derives from active surveillance for KD cases hospitalized at Rady Children’s Hospital San Diego, a tertiary referral center that cares for more than 90% of the acute KD cases in San Diego County.

Time series are formed by KD patients, the date being that of hospital admission recorded for all subjects. For subjects with multiple admissions, only the first hospitalization date was used. To provide uniform assessment among sites, the date of admission was used as a surrogate for the date of fever onset, even though for comparison with other studies, the average day of KD diagnosis is normally centered on the 5th day of fever. However, analyses were reproduced with calculated dates of fever onset and results did not differ (p < 0.001, t-test).

Filtering of the interannual component of KD time series

An eigendecomposition analysis applied to the data covariance matrix of monthly KD time series in Japan and San Diego was used to partition signal contributions by frequency, with the aid of adaptative nonparametric functions (Elsner and Tsonis 1996). An embedding dimension or order of the decomposition of 40 was selected, as it allows a proper characterization of those signals with period larger than the annual cycle (Dettinger et al. 1995). In this study, only principal components coming from significant eigenvalues that captured interannual frequencies were kept for further reconstruction (black line in Figures 1a,d). For comparison, a data tapering window procedure to minimize spectral leakage was applied to each series to remove spurious results and maximize both the decomposition and the reconstruction. Find a complete description of the data sources and further details of the filtering procedure in Rodó et al. (2011).

Significance level of composite anomalies

The significance level of the interannual ocean and atmosphere anomaly maps (Figure 2 and Supplementary Figure 4) was computed using a bootstrap method (Efron and Tibshirani 1993), which was especially designed to evaluate data whose degrees of freedom are reduced by low-pass filters and the averaging of small samples of events in the composites. The reference probability distribution function was locally computed by means of 10000 white-noise time series, which were then tested against an autoregressive process of order 1 having the same mean, standard deviation, and temporal one-lag autocorrelation value as the interannual grid-point time series of the variable tested. Find further details in Ballester et al. (2011).

References

Ballester J, Rodríguez-Arias MA, Rodó X. A new extratropical tracer describing the role of the western Pacific in the onset of El Niño: Implications for ENSO understanding and forecasting. Journal of Climate 24, 1425-1437 (2011).

Dettinger MD, Strong CM, Weibel W, Ghil M, Yiou P. Software for singular spectrum analysis of noisy time series. Eos, Transactions, American Geophysical Union 76, 12 (1995).

Efron B, Tibshirani R. Introduction to the Bootstrap. Chapman and Hall, 436 pp (1993).

Elsner JB, Tsonis AA. Singular Spectrum Analysis: A New Tool in Time Series Analysis. Plenum Press, 164 pp (1996).

Rodó X, Ballester J, Cayan D, Melish ME, Nakamura Y, Uehara R, Burns JC. Association of Kawasaki disease with tropospheric wind patterns. Nature Scientific Reports 1, 152 (2011).

Figure Captions

Supplementary Figure 1. Mean seasonal cycle of surface (a-d), 700hPa (e-h) and 300hPa (i-l) winds (m/s) for 1994-2008.

Arrows correspond to horizontal winds, and the shading to the zonal component only. The green box depicts the zonal path connecting the Asian continent with Japan and the western coast of the United States, which is strongest at all vertical levels in boreal winter (DJF).

Supplementary Figure 2. Major epidemics of KD in Japan (1978/79, 1981/82, 1985/86).

Panels a,d,g and c,f,i depict the time average of surface winds (arrows, m/s) and the change in KD at the prefecture level (colored dots, cases per million) for the increasing (autumn and winter) and decreasing (spring and summer) phases of the epidemics shown in red and blue in panels b,e,h, respectively.

Supplementary Figure 3. Ocean configuration for the individual interannual peaks of KD in Japan (a-c) and San Diego (d-f).

Panels a and d depict the correspondence between equatorial surface ocean warming and the standardized (unitless) interannual component of KD for the peaks in Figures 1a,d for Japan and San Diego, respectively. Equatorial warming is here depicted as the longitudinal range (horizontal lines) and maximum (dots) of positive standardized interannual anomalies of equatorially-averaged (5S-5N) sea surface temperature. From left to right, map columns show the standardized (unitless) interannual anomalies of sea surface temperature for the peaks shown in red and pink in Figure 1a for Japan, and in red and pink in Figure 1d for San Diego.

Supplementary Figure 4. Ocean-atmosphere mean composites for the interannual peaks of KD in Japan (a-f) and San Diego (g-r).

From left to right, columns show the standardized (unitless) interannual anomalies for the peaks shown in red and pink in Figure 1a for Japan, and in red and pink in Figure 1d for San Diego. See the individual events in Supplementary Figures 3b,c, 4b,c, respectively. Maps correspond to the composites for sea surface temperature (shading and contours in a,d,g,m), 500hPa geopotential height (shading and contours in j,p) and 500hPa winds (arrows in j,p). White contours depict shaded areas in the maps with significant anomalies (p < 0.05). Composite anomalies of tropospheric winds along zonal (b,e,h,k,l,n,q,r) and meridional (c,f,i,o) cross sections (see green and brown boxes in the corresponding column map) are depicted in the two bottom rows, in which the shading corresponds to the meridional (positive means northwards) or zonal (positive means eastwards) component, respectively.

Supplementary Figure 5. Same as Figures 2e-j for the ocean-atmosphere composites in San Diego, but using the same criterion for the selection of the peaks than in Rodó et al. (2011).

The seven peaks reaching the +0.75 standard deviation criterion in the reconstructed interannual component were included in this case in the ocean-atmosphere composites. Two events were thus added: the red peak in 1994/95 (KD = +0.79σ) and the pink peak in 2005/06 (KD = +0.98σ). This figure shows that results are not sensitive to the choice of the threshold.

4