The cross-equatorial flow over the western Indian Ocean, known as the Findlater Jet, plays an important role in the monsoonal circulation of the region. During the boreal summer, there is southerly flow across the equator that is concentrated along the East African highlands. During the boreal winter, there is a reversal in wind direction across the equator. Madagascar, the world's fourth largest island, with heights in excess of 1 km represents a significant obstacle to the flow whose impact on this jet has not been fully characterized. Here we use diagnostic tools developed to investigate atmospheric flow distortion by Greenland's high topography to study this interaction. We show that there is a bifurcation of the Findlater Jet by Madagascar during the boreal summer and localized tip jets at the island's northern and southern ends. During the boreal winter, the northern tip jet reverses direction and weakens, while the southern tip jet maintains its direction and magnitude. We show that rotational effects are important for these interactions but not dominant and result in an enhancement of the northern tip jet; while allowing for existence of the southern tip jet. As will also be shown, this flow distortion has impacts on the meteorology and oceanography of the region including the forcing of oceanic eddies in the Mozambique Channel, a modulation of the southward displacement of the Inter-Tropical Convergence Zone (ITCZ) and a splitting of the boreal summer cross-equatorial mass transport associated with the Findlater Jet into two branches.
 The climate of the Indian Ocean is dominated by the annual reversal in the direction of the surface wind field that is most pronounced along its western boundary with Africa where the flow is referred to as the Findlater or Somali Jet [Findlater, 1969]. The East African highlands, with an average elevation on the order of 1 km, deflect the easterly flow over the southern Indian Ocean northward during the boreal summer resulting in a narrow core of high winds, with a width of several hundred kilometers, which is an important characteristic of the Findlater Jet [Findlater, 1969; Hoskins and Rodwell, 1995]. Indeed during the boreal summer, this narrow jet is responsible for approximately 50% of the global tropospheric cross-equatorial mass flux [Boos and Emanuel, 2009].
 Although the role that the East African highlands plays in the structure of the Findlater Jet has been recognized for some time, the impact of the high topography of Madagascar has not been extensively studied. In the original work on the Findlater Jet, a localized region of elevated wind speed was observed just downstream of the northern tip of Madagascar [Findlater, 1969]. Krishnamurti et al.  used a single-level primitive equation model to show that this downstream maximum was associated with the high topography of Madagascar; however, no mechanism for its development was presented. Collins et al.  used QuikSCAT scatterometer data and reanalysis products to identify a number of topographically forced features of the surface wind field over the western Indian Ocean including local acceleration near the northern and southern tip of Madagascar. Collins et al.  also highlighted the important role that these features of the surface wind field play in forcing the oceanic eddies in the Mozambique Channel.
 The flow distortion associated with the interaction of high topography with the larger-scale atmospheric circulation has been extensively studied in southern Greenland [Renfrew et al., 2008]. Cape Farewell, Greenland's southern most point, is the windiest location on the ocean's surface [Sampe and Xie, 2007]. In addition, Cape Farewell is unique in that the high wind speeds, known as tip jets, can be either westerly or easterly in nature [Moore, 2003]. The dynamics of these two classes of Cape Farewell tip jets are very different [Doyle and Shapiro, 1999; Moore and Renfrew, 2005]. In particular, for easterly tip jets the ageostrophic acceleration resulting from the flow impinging on the topographic barrier contributes to their formation [Renfrew et al., 2009]. In this sense, easterly Cape Farewell tip jets are examples of so-called corner jets that in the Northern Hemisphere are enhanced to the left of the barrier [Smith, 1982; Barstad and Gronas, 2005]. Moore  introduced a new diagnostic, obtained through the partitioning of the occurrence frequency of high-speed winds by direction, which allowed for a more complete view of the impact that the high topography of Greenland has on the surface wind field and in so doing identified new regions along its coasts where high-speed left-handed corner jets occur.
 In this paper, we will use the techniques developed to diagnose the characteristics of the high-speed winds near Greenland to investigate the impact that the high topography of Madagascar has on the structure of the Findlater Jet.
2 Data and Methods
 We use 10 m winds over the ocean derived from the scatterometer on the QuikSCAT satellite that have been obtained using a revised geophysical model function that improves the retrievals at high wind speeds [Meissner and Wentz, 2012]. The data are available twice daily at a horizontal resolution of 0.25° for the period from July 1999 to November 2009. We also make use of the 3-D wind, 10 m wind and the sea-level pressure fields from the Interim Reanalysis from the European Centre for Medium-Range Weather Forecasts (ECMWF) (ERA-I) [Dee et al., 2011]. We make use the 0.75° 6 hourly fields for the period from January 1979 to August 2012. The ERA-I employs a terrain following vertical coordinate and we use the model level data in this analysis. Typically there are 14 model levels in the lowest 1.5 km of the atmosphere.
 A comparison of the QuikSCAT and ERA-I 10 m winds showed good overall agreement over the western tropical Indian Ocean [Collins et al., 2012] with a tendency of the ERA-I to overestimate the magnitude of the wind speed in the region off the Horn of Africa and in the lee of Madagascar where detailed comparisons were undertaken. In addition, Collins et al.  noted that the ERA-I was better able to resolve fine-scale structures in the surface wind field forced by topography as compared to the coarser NCEP Reanalysis.
 The QuikSCAT and ERA-I data were averaged into seasonal means for both the boreal winter (defined to be January and February) and boreal summer (defined to be July and August) periods. In what follows, we will refer to these periods as “winter” and “summer”, respectively. These periods were selected because they represent the core periods for which one has either northerly flow (winter) or southerly flow (summer) in the Findlater Jet [Boos and Emanuel, 2009; Collins et al., 2012]. In addition, the frequency of occurrence of 10 m westerly and easterly winds with wind speeds in excess of 10 m/s were also calculated from the ERA-I [Moore, 2012].
 Figure 1 shows the summer and winter mean 10 m wind fields from the QuikSCAT and ERA-I data sets. Both data sets are able to capture the overall structure of the Findlater Jet with its southerly cross-equatorial transport during the summer that is focused along the East African highlands with a reversal in cross-equatorial flow during the winter. During the winter, there are still easterlies to the south of the equator and the interaction of these two counter-flowing wind systems results in a minimum in wind speed that is associated with the ITCZ [Bigg, 1992]. During the summer, both data sets capture the local maximum in the magnitude of the 10m wind just to the north of Madagascar that was identified in Findlater . In addition, there is a local maximum in the magnitude of the 10m wind along the southern tip of Madagascar in both data sets that is most pronounced during the winter. Furthermore during the winter, there is an anticyclonic circulation to the northwest of Madagascar over the northern Mozambique Chanel.
 There are some features of the surface flow in the vicinity of Madagascar that are better resolved by the ERA-I as compared to QuikSCAT as a result of the latter's problem with coastal contamination [Owen and Long, 2009]. This allows for the ERA-I's more complete representation of the interaction of the Findlater Jet with Madagascar. During the summer, this interaction includes a bifurcation of the jet into northward and southward branches upwind of Madagascar with a region of deceleration on the upwind side of the island. The southern branch of the jet turns northward along the Mozambique Channel before merging with the main branch north of the island.
 For the period of overlap between the two data sets, 2000 to 2009, the RMS error between the magnitude of both the summer and winter mean climatological 10 m wind over the domain shown in Figure 1 is on the order of 1 m/s. This error is comparable to that between QuikSCAT and buoy winds in the region [Collins et al., 2012].
 Figure 2 shows the summer and winter mean sea-level pressure fields from the ERA-I. During the summer, the pronounced cross-equatorial pressure gradient and Mascarene High are evident [Krishnamurti and Bhalme, 1976] as is the localized ridging on the upwind side of Madagascar extending eastward to 50°E and troughing in its lee [Smith, 1982]. The situation during the winter is more complex with a minimum in pressure south of the equator that is associated with the ITCZ. The pressure anomaly associated with the interaction of the easterly flow with Madagascar results in a gradient across the island that lead to a pronounced perturbation in the pressure field that extends across the Mozambique Channel.
 The frequency of occurrence of easterly and westerly flow with the magnitude of the 10 m wind speed exceeding 10 m/s during the summer and winter are shown in Figure 3. With respect to easterly flow during the summer (Figure 3a), the northern and southern tips of Madagascar can be seen to be regions where the occurrence frequency is elevated with the northern region having a significantly larger magnitude. In the lee of Madagascar, there are two regions where the occurrence frequency is elevated; one of which is located in the saddle between the two wind speed nadirs (Figures 1a and 1c). The other is situated along the East African coast. During the winter, an elevated frequency of occurrence of easterly flow again occurs near the southern tip of Madagascar and along the coast of the Horn of Africa north of the equator (Figure 3b).
 During the summer, there is an elevated occurrence of westerly flow associated with the flow along the coast of the Horn of Africa (Figure 3c). Along the East African coast there is also a localized region of elevated occurrence frequency of westerly flow that is nearby the region of elevated occurrence frequency of easterly flow (Figure 3a). During the winter, the only region with an elevated occurrence frequency of westerly flow is situated to the east of the northern tip of Madagascar.
 The vertical structure of the Findlater Jet in the vicinity of Madagascar is shown in Figure 4. The meridional cross-section of the summer mean zonal wind along 50°E (Figure 4a) shows the core of the northern tip jet situated near 10°S. The strongest winds in the jet are at height of 1.1km. This is approximately the same height as the maximum winds in the undisturbed flow upstream of Madagascar. There is also evidence of the southern tip jet with a local maximum in easterly winds near 23°E at a height of approximately 0.5 km. Note that this cross-section is situated upwind of the island (Figure 1) and so the presence of the southern jet core is evidence of the upwind effect of Madagascar on the wind field (Figure 2a).
 The meridional cross-section of the winter mean zonal wind along 45°E (Figure 4b) shows the core of the southern tip jet is near 27°S. The core of the jet is at a height of 0.4 km, slightly below the maximum height of the topography in the region (Figure 1). North of Madagascar, there evidence of the anticyclonic circulation noted in the region (Figures 1c and 1d). This feature is quite deep extending to a height of approximately 4 km. The southerly component of this circulation shows an enhancement over the Ethiopian highlands.
 In this paper, the interaction of the high topography of Madagascar with the Findlater Jet has been examined. Although the island is elliptical in shape, its width is important in determining the pressure gradient across the island [Smith, 1982] and so the analysis can be simplified if one assumes a circular island with half-width given by a and height h. The flow in the vicinity of the island is determined by the nondimensional barrier height
and the inverse Rossby number
where N is the Brunt-Väisälä frequency, U is the upwind wind speed and f is the Coriolis parameter [Smith, 1982]. For Madagascar, a is of O(100 km), h is of O(1 km), N is of O(10− 2s− 1), and U is of O(10 ms− 1) resulting in . This places the flow in the regime where wave-breaking over the island may occur [Smith, 1982, 1989]. Madagascar spans approximately 15° of latitude in a region where the Coriolis parameter is a rapidly changing function of latitude resulting in an inverse Rossby number that varies by a factor of 2 from its northern tip where ε is of O(0.25) to its southern tip where ε is of O(0.5). This places the flow in a region where rotation effects are important but not dominant. It should be noted that the high African topography downwind of Madagascar also plays a role in the flow distortion. This is especially true over the Mozambique Channel as can be seen in the sea-level pressure field (Figure 2).
 For low inverse Rossby number flow past an isolated topographic barrier [Smith, 1982], there exists a pressure gradient across the barrier that for the parameters of the flow under study is predicted to be of O(2 mb). The sea-level pressure pattern during the summer shows evidence of a perturbation that is of this magnitude (Figure 2a). During the winter, the reversal of the flow across the island results in a complex quadrupolar pressure perturbation with a positive gradient to the south and a weaker negative gradient to the north (Figure 2b).
 In the irrotational limit, i.e., ε = 0, there is upstream deceleration of the flow with symmetric deflection of air around the barrier [Smith, 1982]. This results in corner jets of equal strength to the north and south of the barrier. With rotation present, the upstream deceleration results in a reduction in the Coriolis Force that results, in the Southern Hemisphere, in a preferential deflection toward the right of the barrier. In the current situation, this would result in an enhancement of the northern tip jet. This is the situation that occurs during the summer (Figures 1a and 1c). When , this deflection is sufficiently weak so that there is still flow to the left of the barrier in the Southern Hemisphere [Smith, 1982]. When , there is northerly flow to the south of the barrier resulting in deflection to the right everywhere in the vicinity of the obstacle. For the Findlater Jet and Madagascar, the inverse Rossby number in the region is such that one would expect, as is observed, to observe flow around the southern tip. A comparison of the magnitude of the northern and southern tip jets shows that, in agreement with this theory, there is an acceleration of the northern jet with respect to the upstream flow; while the southern jet, although well defined as a local region of accelerated flow, has wind speeds that are reduced as compared to the upstream flow.
 The vertical cross-sections through the northern and southern tip jets (Figure 4) show clear evidence of well-defined cores above the surface. For the case of the northern tip jet, the core is at a similar height to that upstream of the topography suggesting that its vertical structure is not determined by the interaction with the barrier. In contrast, the height of the core of the southern jet is at the height of the topography suggesting that its vertical extent is determined by the interaction.
 The splitting of the flow by Madagascar results in equatorward flow through the Mozambique Channel during both the summer and winter. During the summer, the ERA-I data indicate that the tropospheric mass transport through the Mozambique Channel is approximately 30% of that in the branch of the Findlater Jet to the east of Madagascar. Given that the transport associated with the Findlater Jet during the summer is approximately 50% of the total cross-equatorial transport, this suggests that the southerly flow through the Mozambique Channel represents an important and hitherto unrecognized feature of this transport. During the winter, the interaction of this southerly flow with the cross-equatorial northerly flow results in a region of low wind speed along the channel that is associated with the largest southward displacement of the ITCZ anywhere in the tropics [Bigg, 1992]. It therefore follows that any change in the intensity of this southerly flow, and by extension the magnitude of the southerly tip jet, would modulate the meridional displacement of the ITCZ in this region impacting rainfall over Madagascar and East Africa during the boreal winter.
 In addition, the Mozambique Channel is the source of the Agulhas Current, the most intense western boundary current in the Southern Hemisphere, which plays an important role in the climate of Southern Africa and acts as a source of heat and salt to the Atlantic Ocean [Backeberg et al., 2012; Rouault et al., 2009]. There is evidence of the formation of an anticyclonic eddy in the northern section of the channel during the boreal winter that subsequently moves southward advecting subtropical waters into the Agulhas Current [Schouten et al., 2005]. The anticyclonic circulation that forms in this region as a result of the interaction of the northerly cross-equatorial flow with Madagascar during the boreal winter (Figures 1b and 1d) may play a role in forcing this eddy. The southern tip jet may also play a role in upwelling along the South East Madagascar Current and in the eddies that are shed as this current leaves the island's southern coast and merges with the Agulhas Current [de Ruijter et al., 2004].
 In summary, the interaction of the Findlater Jet with Madagascar results in flow distortion that is in agreement with the theory of Smith  for low inverse Rossby number flow. Among the important characteristics of the flow that are in agreement with this theory are the presence of both southern and northern Madagascar tip jets with a preferential enhancement of the flow in the northern jet. There is also evidence of a complex wind speed nadir in the lee of the island that may be evidence of wave-breaking and downslope acceleration of the flow. However, there is a need to generalize this theory to take into account the large variation in the Coriolis parameter along Madagascar, a characteristic of this situation that is not present in the Greenland region. In addition in the Greenland region, it is background transient low-pressure systems that are responsible for the tip jets that form near Cape Farewell [Moore, 2003]. In the case of Madagascar, it is a stationary high-pressure system.
 The presence of the southern jet results in equatorward flow along the Mozambique Channel that has impacts on the meteorology and oceanography of the region. The characteristics of the background environment are however such as to place the flow close to the point where this bifurcation vanishes. There is evidence of a trend toward an intensification of the easterly trade winds over the southern Indian Ocean [Han et al., 2010; Rouault et al., 2009] and such changes would result in a decrease in the inverse Rossby number that may result in changes in the flow splitting by Madagascar.
 The author would like to thank the ECMWF for access to the ERA-I data and Remote Sensing Science for access to the QuikSCAT data. The author would like to thank B. Harden and the reviewers for comments that improved the manuscript. The author was supported by the Natural Sciences and Engineering Research Council of Canada.
 The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.