Interannual to Decadal Variability of Tropical Indian Ocean Sea Surface
Total Page:16
File Type:pdf, Size:1020Kb
Manuscript (non-LaTeX) Click here to download Manuscript (non-LaTeX) manuscript.docx 1 Interannual to Decadal Variability of Tropical 2 Indian Ocean Sea Surface Temperature: 3 Pacific Influence versus Local Internal 4 Variability 5 6 Gang Wang1*, Matthew Newman2, 3 and Weiqing Han1 7 8 1 ATOC, University of Colorado, Boulder, Colorado 9 2 CIRES, University of Colorado, Boulder, Colorado 10 3 NOAA Earth Systems Research Laboratory, Boulder, Colorado 11 12 *Corresponding author address: Gang Wang, University of Colorado Boulder, 311 13 UCB, Boulder, CO 80309 14 15 Submitted to Journal of Climate, 06/14/2018 16 Abstract 17 18 The Indian Ocean has received increasing attention for its large impacts on regional and 19 global climate. However, sea surface temperature (SST) variability arising from Indian 20 Ocean internal processes has not been well understood particularly on decadal and 21 longer timescales, and the external influence from the Tropical Pacific has not been 22 quantified. This paper analyzes the interannual-to-decadal SST variability in the 23 Tropical Indian Ocean in observations and explores the external influence from the 24 Pacific versus internal processes within the Indian Ocean using a Linear Inverse Model 25 (LIM). Coupling between Indian Ocean and tropical Pacific SST anomalies (SSTAs) is 26 assessed both within the LIM dynamical operator and the unpredictable stochastic noise 27 that forces the system. SSTA variance decreases significantly in the Tropical Indian 28 Ocean in the absence of this coupling, especially when the one-way impact from the 29 Pacific to the Indian Ocean is removed. On the other hand, the Indian Ocean also affects 30 the Pacific, making the interaction a complete two-way process. Specifically, the Indian 31 Ocean Basin (IOB) mode, apart from its strong relationship to global mean SST time 32 series, is primarily a response to Pacific ENSO forcing. In contrast, the Indian Ocean 33 Dipole (IOD) and Subtropical Indian Ocean Dipole (SIOD) modes appear to be 34 generated by internal modes of Indian Ocean dynamics whose evolution is altered by 35 coupling with the tropical Pacific. While both modes gain variance from ENSO forcing, 36 the IOD exerts significant control of ENSO timing, and the SIOD is almost in 37 quadrature with the 2-yr ENSO component. 38 1. Introduction 39 40 The Indian Ocean plays an important role in the Earth’s climate on timescales ranging 41 from intraseasonal to multidecadal. Recent studies show that the warming trend and 42 decadal variability of Indian Ocean sea surface temperature (SST) can have large 43 impacts on climate both within the Indian Ocean rim regions and in other sectors of 44 the globe via atmospheric teleconnection (see review of Han et al. 2014a). However, 45 as the Indian Ocean is strongly affected by El Niño–Southern Oscillation (ENSO), the 46 SST variability generated by processes intrinsic to the Indian Ocean is difficult to 47 quantify, and studies on decadal and longer timescale variability are lacking (e.g., see 48 review of Han et al. 2014a). 49 On interannual to decadal timescales, Indian Ocean SST variability is 50 dominated primarily by two patterns: the Indian Ocean Basin mode (IOB) and the 51 Indian Ocean Dipole (IOD). [Hereafter, decadal variability is broadly referred to as 52 variability on a time scale of one to a few decades.] The IOB, the leading empirical 53 orthogonal function (EOF) of Indian Ocean SST interannual variability, has a basin- 54 wide warming/cooling pattern across the tropical Indian Ocean and usually lags 55 ENSO by a few months. Studies have suggested that it is largely driven by ENSO- 56 induced cloud and surface flux variations (e.g., Klein et al., 1999). 57 On decadal time scales, before about 1985 the IOB and the Interdecadal 58 Pacific Oscillation (IPO), an ENSO-like pattern of decadal variability (Power et al., 59 1999) were positively correlated. Since 1985, however, the correlation has been 60 negative (Han et al., 2014b). Recent analysis of climate model experiments suggests 61 that this reversed relationship resulted from the external forcing of anthropogenic 62 greenhouse gases on multi-decadal timescales (Dong and McPhaden, 2017; Zhang et 63 al. 2018) and volcanic eruptions on decadal timescales (Zhang et al. 2018). However, 64 the detailed features of SST variability internal to the Indian Ocean, including the 65 effects of natural internal climate variability and natural external forcing, remain 66 unclear. 67 The IOD, the second EOF of Indian Ocean SST interannual variability, has an 68 east-west SST dipole structure accompanied by easterly wind anomalies (Saji et al., 69 1999; Webster et al., 1999). Empirical studies suggest that while some IOD events co- 70 occur with ENSO, others are independent of ENSO (e.g., Allan et al., 2001; 71 Yamagata et al., 2004; Chang et al., 2006; Meyers et al. 2007; Sun et al. 2015). 72 Climate model simulations agree with the observational analyses, showing that in 73 some models, the IOD tends to be triggered by ENSO (e.g., Yu and Lau, 2005; 74 Loschnigg et al., 2003; Lau and Nath, 2004; Fischer et al., 2005; Saji et al. 2006), 75 whereas in others, the IOD can be self-generated and the dominant SST pattern is 76 unchanged when the Pacific Ocean is decoupled. On decadal timescales, variations of 77 the IOD index are independent of decadal variability of ENSO (Song et al. 2007; 78 Tozuka et al. 2007), suggesting that decadal variability of the IOD may be intrinsic to 79 the Indian Ocean ocean-atmospheric coupled system. 80 In addition to the IOB and IOD, other SST patterns of interannual variability, 81 such as the Subtropical Indian Ocean Dipole (SIOD; e.g., Behera and Yamagata, 82 2001; Reason, 2002; Suzuki et al. 2004), have been identified. The SIOD is located in 83 the extratropical South Indian Ocean and its formation appears related to the surface 84 mixed layer heat anomalies caused by heat flux variations (Hermes and Reason 2005; 85 Huang and Shukla, 2007; Morioka et al., 2010, 2013; Kataoka et al., 2012). It has also 86 been associated with ENSO (e.g., Zinke et al. 2004; Hermes and Reason 2005). 87 While the atmospheric bridge facilitates the interbasin interaction between the 88 Indian and Pacific Oceans (e.g., Alexander et al., 2002; Izumo et al., 2014), the 89 Pacific can also affect the Indian Ocean through an oceanic connection: the 90 Indonesian Throughflow (ITF). The ITF transports warmer and fresher waters from 91 the Pacific to the Indian Ocean (e.g., Gordon and Fine, 1996; Meyers 1996; van 92 Sebille et al., 2014). As a result, the Pacific impact may be communicated to the 93 Indian Ocean in a variety of ways. Yet, the overall impact of ENSO on the Indian 94 Ocean variability is still not completely clear. On the other hand, since most previous 95 studies focus on the relationship between ENSO and individual climate modes in the 96 Indian Ocean, the overall Indian Ocean internal variation and its response to the 97 Pacific variability is still missing and worth deeper investigation. 98 To quantify the coupled dynamics between the tropical Pacific and Indian 99 oceans, in this study we determine a Linear Inverse Model (LIM) from the observed 100 evolution of seasonal SST anomalies within each basin. We use the LIM to diagnose 101 the relative importance of internal Indian Ocean dynamics compared to its coupling 102 with the Tropical Pacific for interannual-to-decadal SST variability, with a focus on 103 the leading Indian ocean SST EOF patterns (IOB, IOD and SIOD). The effect of 104 anthropogenic warming is removed prior to our analysis. The rest of the paper is 105 organized as follows: Section 2 briefly introduces the LIM, section 3 presents data 106 and dominant patterns in the Tropical Indian Ocean, section 4 explores the leading 107 eigenmodes on the basis of LIM application, section 5 reports the primary results, and 108 finally, section 6 provides a summary and discussion. 109 110 2. Linear Inverse Model 111 112 Here we apply LIM to a state vector made up of tropical Pacific and Indian Ocean 113 SST anomalies (SSTA). LIM empirically estimates the linear dynamics of a system 114 from its time-lag covariance statistics. It has been widely used in the climate science 115 community, from studying ENSO dynamics to decadal climate prediction (e.g., 116 Penland and Sardeshmukh 1995; Newman and Sardeshmukh, 2003; Newman 2007; 117 Newman and Sardeshmukh, 2008; Solomon and Newman, 2012; Cavanaugh et al., 118 2015; Newman et al., 2016). LIM can extract dynamically relevant coupled structures 119 that oscillate at different time scales without time filtering (e.g., Penland and 120 Sardeshmukh 1995; Newman 2007), and it can provide insight into coupling between 121 different processes or different domains (e.g., Newman 2007; Newman et al. 2011). 122 The LIM is briefly summarized below. Here, x denotes SSTA, and the linear 123 relationship can be written as 124 ��/�� = ��(�) + �(�) (1) 125 where �(�) is unpredictable white noise (in time, not necessarily in space) and L is 126 the dynamical operator. Note that (1) can be a suitable -- and importantly, testable -- 127 approximation of a highly nonlinear system whose nonlinear terms decorrelate much 128 more rapidly than its linear terms (e.g., Hasselmann 1976; Penland 1996; Just et al. 129 2001). Here, this means that tropical Pacific-Indian ocean coupled processes acting on 130 time scales shorter than about a season will be represented by the noise in (1) and not 131 by the deterministic dynamics. 132 The forward solution of (1), 133 �(� + �) = exp(��)�(�) + � = �(�)�(�) + � , (2) 134 describes the SSTA evolution, where �(�) = exp(��) and ε is the LIM forecast error.