Solar warming of the south-central Pacific

Surface solar radiation is found to have contributed significantly and positively to the record warming event in the south-central Pacific (SCP) that peaked in December of 2009. The SCP region is within a positive teleconnection pattern between sea surface temperature anomalies in the equatorial Pacific and basin-wide surface solar radiation, as revealed by a 24-year time series; the pattern extends southeast from the western equatorial Pacific toward the SCP region. The results are consistent with the ‘atmospheric bridge postulation’ on El Niño teleconnection with extratropical sea surface temperature anomalies, but with the extension to cloud cover and surface solar radiation over the mid-latitude southern oceans.


Introduction
A record warming event occurred in the south-central Pacific (SCP) starting in September and peaking in December of 2009. Lee et al. (2010) established this event as one of the most intense warming events of the past two decades and attributed the warming to two factors with comparable effects: the reduction in evaporative cooling and Ekman advection of cold water from the south. They postulate that evaporative reduction is caused by the negative anomalies in wind speed and westerly wind component of the anticyclone that persisted through October and November over the region, as related to the 2009 El Niño. Boening, Lee, and Zlotnicki (2011) attribute the observed high ocean bottom pressure and sea surface height in the same event to mass convergence in the ocean resulting from wind-stress curl associated with the anticyclone. Lee et al. (2010) examined various components of surface heat fluxes and concluded that the contribution of solar heating was not important. They used the re-analysis product of the National Centers for Environmental Prediction (NCEP) (Kanamitsu et al. 2002), which has shown reduced solar heating during the warming event of 2009. The reduction in solar heating runs against the common conception that clear-sky conditions associated with an anticyclone will let in more solar flux. The accuracy of the downwelling surface shortwave radiative fluxes (SSWR) from operational numerical weather prediction models has been known to be problematic (e.g. Kanamitsu et al. 2002), primarily due to uncertainties in cloud parameterization. Using space-based observations, the significance of solar heating relative to evaporative cooling in driving sea surface temperature (SST) change was previously demonstrated by Liu and Gautier (1990) in tropical oceans and by Liu, Zheng, and Bishop (1994) over global oceans. A re-examination of the contribution of SSWR to the 2009 heating event will be presented here. A data set of SSWR based on observations from the Moderate Resolution Imaging Spectroradiometer (MODIS) on Terra and Aqua missions Platnick et al. 2003), using an algorithm described by Wang and Pinker (2009), was produced for this study for the period of July 2006 to June 2010.

SST
Many SST data sets are available. We followed Lee et al. (2010), who used Level 4 from the Advanced Very High Resolution Radiometer (AVHRR) monthly product at ¼ resolution (Reynolds et al. 2007), as produced by the Group for High Resolution SST (GHRSST), to study the warming event at SCP and to calculate the El Niño Southern Oscillation (ENSO) indices for the period 1982-2010. For the warming event discussed in Section 3, SST anomalies are computed by subtracting the monthly mean of the previous three years from the 2009 values; for computing correlation with the El Niño index discussed in Section 4, the anomalies are computed by subtracting the climatology compiled from 29 years of data (August 1982-July 2010) from the 2009 values.

SSWR
This study requires information on both SST and SSWR at long timescales (longest available) that extend to mid-2010. The International Satellite Cloud Climatology Project (ISCCP) has been collecting and calibrating the infrared and visible radiances obtained from imaging radiometers carried on the international constellation of weather satellites since July 1983 (Rossow and Schiffer 1999). These observations have been widely used for estimating SSWR (Zhang et al. 2004;Gupta et al. 1999;Hatzianastassiou et al. 2007;Ma and Pinker 2012) because of the frequent sampling from geostationary sensors. The product that has been produced at the NASA Goddard Space Institute and is known as ISCCP-FD (Zhang et al. 2004) will be used in this study. The datasets are available at three-hourly intervals and 280 km spatial resolution over the whole globe between 1983 and 2007. The 24 years of ISCCP-FD SSWR are used to examine the El Niño teleconnections in Section 4. Inter-annual anomalies of SSWR from ISSCP-FD are computed by subtracting climatological mean annual cycles compiled from 24 years of data (January 1984-December 2007. To cover the warming event of 2009 in the SCP, four years of SSWR from July 2006 to June 2010, covering the Pacific Ocean south of 20°N, were produced for this study, based on observations from MODIS on board both NASA Terra and Aqua satellites in complementary orbits, using the algorithms of Pinker and Wang (2003) and Wang and Pinker (2009). The algorithm has been extensively evaluated over land using the Baseline Surface Radiation Network (BSRN) observations. A special effort has been made to evaluate the SSWR fluxes over oceans, using measurements from the Pilot Research Moored Array in the Tropical Atlantic (PIRATA) moorings (Bourlès et al. 2008), and from the Tropical Atmosphere Ocean/Triangle Trans-Ocean Buoy Network (TAO/ TRITON) moorings in the tropical Pacific Ocean (McPhaden et al. 1998). Observations were also used from special experiments such as the NOAA Kuroshio Extension Observatory (KEO) moored buoy, the JAMSTEC Kuroshio Extension Observatory (JKEO) moored buoy, and the CLIVAR Mode Water Dynamic Experiment (CLIMODE) buoys as described in Pinker, Wang, and Grodsky (2009) and Niu, Pinker, and Cronin (2010). For the monthly data used in this study, the bias was estimated to be 2 and 1 W m -2 , and the root-mean-square error (RMSE) is 13 and 11 W m -2 , for the Atlantic and Pacific oceans, respectively; the RMSEs are about 5% of the observed means. The SSWR anomalies from July 2009 to June 2010 were computed by subtracting the monthly mean of the previous three years. -5°N) are outlined. The anomalies in the SCP region reach 2.1°C, and are higher than the 1.5°C anomalies in Niño4. Figure 2 shows that SSWR anomalies can reach 23 W m -2 ; there is strong solar warming of the ocean, in contradiction to the cooling in the NCEP results used in the earlier study by Lee et al. (2010). In Figure 3, SSWR rises with SST (at both SCP and Niño4) in October 2009 and reaches near peak value with SST in December 2009-January 2010. SST anomalies at SCP appear to drop faster than SSWR anomalies after January 2010. Table 1 shows that the anomalous SSWR warming of the ocean (turns positive) started in November 2009 at 9 W m -2 , increased to 27.7 W m -2 in January 2010, and remained positive until May 2010. Assuming a mixed layer of 50 m depth, following Lee et al. (2010), this magnitude of the SSWR anomaly would have caused a mixed-layer temperature increase of up to 18.8% in January and 56.8% of the SST anomaly in February 2010. Rather than looking at individual months, Lee et al. (2010) estimate the effect of latent heat flux accumulated for three previous months to be 40%, and they also attribute 40% of the warming to ocean dynamic processes. The remaining 20% of unaccounted heating may be caused by SSWR. While SSWR did not contribute significantly to the developing phase of the warming event (October 2009), it made a significant contribution during the peak of the event (December 2009). Moreover, it helped to sustain the warming event   during its decay phase (January-March 2010). Without considering horizontal transport, such analysis only serves as a rough indication of the significant roles of surface heat fluxes.

El Niño teleconnection
The distribution of remote correlation coefficient (contemporary) between SSWR anomalies from ISCCP-FD and SST anomalies at Niño4 and Niño3 (150°W-90°W, 5°S-5°N) for the 24-year period is described in Section 2. A statistical test was performed, and only correlations meeting the 99% significant level are shown in Figure 4. Negative correlation in the equatorial Pacific reflects increased cloudiness with locally positive SST anomalies  Figure 4. Remote contemporary correlation coefficient between (a) Niño3 and (b) Niño4 SST anomalies and SSWR anomalies. (Pavlakis et al. 2008). In the southern Pacific, the pattern of positive correlation is similar for Niño3 and Niño4: they stretch from the western equatorial Pacific to the SCP region. Positive Niño3 and Niño4 SST anomalies (El Niño) correspond to positive input of SSWR into the ocean at SCP. The observed SSWR anomalies during the SCP warming event are part of the El Niño teleconnection and atmospheric bridge (see Section 5). The time series in Figure 5 show that, for the 24-year period examined, large El Niño/La Niña episodes as indicated by Niño3 and Niño4 are associated with warming and cooling in the SCP region. For 288 pairs of monthly data, the correlation coefficients are 0.5 and 0.6 for Niño3 and Niño4, respectively. Strong positive SST anomalies in Niño3 in 1987 and 1997 El Niño episodes are associated with high SSWR in SCP. The La Niña episodes at Niño3 in 1989 and 1999 are associated with negative SSWR anomalies. The December 2009 event is one manifestation of this teleconnection.
The 2009 El Niño episode is exhibited in Niño4 more than in Niño3. In the eastern equatorial Pacific (Niño3), the correlation between local SST and SSWR was weak. The difference in SSWR anomalies is consistent with recent discussions on the characteristics of the central Pacific versus the eastern Pacific El Niño (e.g. Ashok and Yamagata 2009;Kao and Yu 2009) . Figures 4 and 5 show that both eastern and central Pacific SST anomalies have significant correlations with SCP SSWR anomalies. Lee et al. (2010) postulated that the mid-latitude SST anomalies are driven by surface turbulent flux of sensible/latent heat and wind-driven ocean surface Ekman transport, but not by SSWR. We showed that SSWR fluxes, as estimated from MODIS observations, may also contribute positively to the warming anomalies in SCP. There has been a long tradition (e.g. Gill and Niiler 1973;Frankignoul 1985) of parameterizing ocean surface turbulent flux and Ekman transport in terms of atmospheric wind, temperature, and humidity; such data are available from ship-based measurements and operational numerical weather prediction (NWP) models. Surface solar radiation is also affected by atmospheric circulation through cloud cover, but its effect on mid-latitude SST variation is often overlooked and assumed to be negligible because clouds have a small and rapid variability; they are sparsely measured and poorly simulated by NWP. Our results show that SSWR is also a significant factor, particularly in this warming event.

Discussion/Conclusion
A large number of investigations linked SST anomalies in the equatorial Pacific with those over global oceans using observation (e.g. Mestas-Nuñez and Enfield 1999; Garreaud and Battisti 1999). Over mid-latitude oceans the local lapse rate is usually assumed to be too weak to sustain deep convection for SST anomalies to affect largescale atmospheric circulation, except over ocean fronts (Liu, Xie, and Niiler 2007). In the tropical Pacific, SST anomalies during El Niño cause deep convection that affects largescale atmospheric circulation. The El Niño-induced large-scale atmospheric teleconnection alters the near-surface air temperature, humidity, and wind far from the equatorial Pacific. The resulting variation in heat, water, and momentum flux drives local SST anomalies. This 'atmospheric bridge hypothesis' has been reviewed by Kushnir et al. (2002) and Alexander et al. (2002). The results of Lee et al. (2010) are consistent with this hypothesis. Our results (Figure 4) extend the 'atmospheric bridge postulation' to midlatitude cloud cover and surface SSWR. Lee et al. (2010) suggested that central Pacific equatorial warming may be more important than eastern equatorial warming in causing the SCP warming events. This is true for the 2009-2010 event, when the SST anomalies in Niño4 were stronger than in Niño3. Over the long-term, as demonstrated in Figures 4  and 5, there is no significant difference between correlations of Niño3 and Niño4.

Funding
This study was performed at the Jet Propulsion Laboratory, California Institute of Technology, under contract from the National Aeronautics and Space Administration (NASA). It was jointly supported by the NASA Physical Oceanography and the NASA Energy and Water Studies (NEWS) programmes. The work of R.T. Pinker was supported by a collaborative research [grant number ATM-0631792] of the National Science Foundation and benefited from NASA support [grant number NNX08AN40A] from the NASA Science Mission Directorate Division of Earth Science to the University of Maryland.