The evolution of solar wind speed distribution during the period 1985–2008, which covers two solar cycles (22 and 23), has been investigated using multistation interplanetary scintillation (IPS) measurements at 327 MHz. The results obtained here clearly demonstrate that fast (slow) wind areas increase (decrease) systematically as the solar activity diminishes, reaching the maximum (minimum) value at the minimum phase. The intermediate-speed wind areas appear to remain constant over the solar cycle. The preponderance of slow wind at low latitudes was confirmed from our IPS observations throughout the period, while a slight increase in the fast wind area was revealed in the declining to minimum phases. In contrast, the high-latitude solar wind was mostly dominated by the fast wind except for a few years around the maxima. An important point to note is the clear difference in the solar wind speed distribution between the 1996 and 2008 minima. The fast wind areas in 2008 showed a marked increase at low latitudes, which is consistent with in situ observations at 1 AU, and a distinct decrease at high latitudes, resulting in a net decrease at all latitudes, as compared with those in 1996. This difference is ascribed to the weaker polar fields during the 2008 minimum. An excellent positive (negative) correlation between fast (slow) wind areas and the polar fields is revealed from a comparison between IPS and magnetograph observations. The results obtained here suggest a strong control of the Sun's polar field in determining the solar wind acceleration and structure.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Multistation observations of interplanetary scintillation (IPS), which arises from the diffraction of radio waves from a compact source by density fluctuations in the heliosphere, enable to measure the solar wind speed at all latitudes and longitudes [e.g., Dennison and Hewish, 1967]. Utilizing this IPS observation capability, the global distribution of solar wind speed has been investigated extensively since the early 1970s. As result, the large-scale solar wind structure was found to change systematically with the phase of solar activity (see the review of Kojima and Kakinuma ). The solar cycle change in solar wind structure was first reported from three-station 74 MHz IPS observations during 1972–1979 [Coles et al., 1980], and was studied further from more extensive IPS observations covering the period of 1973–1987 [Kojima and Kakinuma, 1987, 1990]. The IPS data clearly demonstrated that the wind speed at high latitudes was considerably faster than that at the solar equator during the declining and minimum phases, and showed that it was dramatically reduced during the maximum phase. In addition, the low-speed regions were found to be distributed along the equator during low solar activity, but were observed to evolve to a more uniform distribution on the source surface as the solar activity rose. A close association between low-speed winds and magnetic neutral lines was also revealed from these IPS observations. The relationships among wind speed, density and angular separation from the magnetic neutral sheet were investigated for the period 1972–1987 by combining IPS observations with coronagraph and magnetograph observations [Rickett and Coles, 1991].
 Such a solar cycle change in the solar wind structure has been repeatedly observed throughout the subsequent cycle (22) and for the early part of cycle 23. However, a quite different feature of the large-scale solar wind structure was clearly revealed for 2008, which corresponded to the late declining phase of the cycle 23 or the minimum phase between cycles 23 and 24, from IPS observations at the Solar-Terrestrial Environment Laboratory (STEL) of Nagoya University [Tokumaru et al., 2009]. The most significant difference in the 2008 observations was a marked increase in solar wind speeds at the equator as well as at the north and south poles. This fact is regarded as a consequence of the weak polar field in the cycle 23/24 minimum. Some unusual aspects of the solar wind in this minimum have been reported from in situ observations, as discussed in section 4 [McComas et al., 2008; Issautier et al., 2008; Smith and Balogh, 2008; Lee et al., 2009]. Thus, it is of great interest to clarify the solar wind evolution for the entire period of cycle 23, and to compare it with that of the previous cycle.
 In this study, we analyzed STEL IPS observations for the period 1985–2008 to elucidate the solar cycle variations in the global distribution of solar wind speed during cycles 22 and 23. Here, the computer-assisted tomography (CAT) method [Kojima et al., 1998, 2007] was used to deconvolve the IPS data. Special attention was paid to peculiarities in the large-scale solar wind structure observed during the late declining phase to the following minimum of cycle 23. The outline of this paper is as follows. In section 2, we briefly describe our IPS observations and the analysis used in this study. In section 3, we show the long-term evolution of the solar wind speed distribution during 1985–2008. We also present a comparison between the 1996 and 2008 IPS data, as well as the relationship between solar wind speeds and polar fields. In section 4, we discuss a distinguishing feature of the solar wind structure observed in the declining phase to minimum of cycle 23 by comparison with in situ measurements. In section 5, we summarize the results of this study.
2. STEL IPS Observations
 IPS measurements of solar wind speed have been carried out regularly since the early 1970s at the Research Institute of Atmospherics (RIA) or at the Solar-Terrestrial Environment Laboratory (STEL; after 1990) of Nagoya University using a multistation system. The initial version of the multistation IPS system consisted of three antennas at Toyokawa, Fuji and Sugadaira, which operated at the VHF frequency of 69 MHz. That version was replaced in 1983 with an advanced three-station system with an operating frequency of 327 MHz, which enabled observations of the solar wind much closer to the Sun (up to 0.1 AU). In 1994, a newly developed antenna for IPS observations was added at Kiso to form a four-station system. The IPS observations using the four-station system were useful, since they provided more robust estimates of the solar wind speed than those of the three-station system owing to a redundancy in the baseline geometry. The four-station system was in operation until 2005 November, when the Toyokawa antenna was closed for an upgrade to a larger one. Since then, IPS observations using the remaining three stations have been conducted regularly to collect solar wind speed data. In the present study, we analyzed those data obtained from 327 MHz IPS observations covering 24 years between 1985 and 2008, which included the entire period of sunspot cycles 22 and 23. That period proved suitable for analyzing the long-term variations of the solar wind from IPS observations, since the solar wind data were collected by using basically the same system and method. We note that changes in the number of observation stations resulted in no significant bias in the IPS data and have tested this by analyzing data for 1996 from only three antenna systems; Fuji, Kiso, and Sugadaira. This analysis using three systems gives results nearly identical with those from the four antenna systems. Since the flow direction was assumed to be radial in our analysis, three stations were sufficient for determining the solar wind speed and its estimation error. The use of four stations did not greatly affect the accuracy of the wind speed measurement, but simply improved the efficiency of the system as a whole, as mentioned above.
 STEL IPS data were collected on a daily basis between April and December for 30–40 lines of sight for selected scintillating radio sources (No IPS data were available during winter owing to interruptions of the antenna operation by snow). The solar wind speeds were estimated by detecting a time delay between spatially separated stations based on the cross correlation analysis of IPS data. Since the IPS observations are the line-of-sight integrations of solar wind speed weighted by density turbulence and the Fresnel filter, we have deconvolved the IPS data using the CAT method [e.g., Kojima et al., 1998]. The IPS CAT method employed here is time sequence tomography [Fujiki et al., 2003; Kojima et al., 2007]. In this method, a large synoptic grid of latitude versus Carrington longitude spanning multiple solar rotations is used as the surface onto which lines of sight are projected without being disconnected at the boundary longitude between adjacent rotations. Such LOS projection is suitable for retrieving the solar wind structure evolving from rotation to rotation as far as it is quasi-stable for a few weeks. The source surface at 2.5 solar radii (RS) is used as a reference sphere onto which the LOS's are projected. A more detailed description of the time sequence tomography has been presented by Kojima et al. . The dimension of the latitude versus the Carrington longitude grid used in our analysis is 180° × 3960° (i.e., 11 rotations). The accuracy of the CAT analysis has been examined by comparing with Ulysses in situ observations [Fujiki et al., 2003; Kojima et al., 2007] and also by making model calculations [Kojima et al., 2004]. From these analyses, it has been shown that the IPS CAT method used here has sufficient reliability and sensitivity for determining solar wind structure. However, these prior studies have also identified the following biases [Kojima et al., 2004]: (1) the IPS CAT analysis tends to underestimate the area of polar fast winds, when they shrink to latitudes higher than 60°. (2) The CAT analysis tends to overestimate the area of the medium-speed (450–700 km/s) region, if the high-speed region is small and the speed gradient between fast and slow winds is sharp. These effects may cause significant errors during solar maximum when the area of coronal holes producing fast winds becomes minimal. Nevertheless, we can safely say that the results obtained from the CAT analysis during solar minimum are sufficiently reliable for the present study.
Figures 1a–1e show solar wind speed maps derived from the time sequence tomography for 1985, 1990, 1996, 2000 and 2008, which correspond to the cycle 21/22 minimum, 22 maximum, 22/23 minimum, 23 maximum, and 23/24 minimum, respectively. In each map, solar wind speed data are plotted in heliographic latitude versus the Carrington rotation number with red (blue) on the map representing low (high) speeds of the solar wind. Note that the Carrington rotation number increases from right to left, whereas the Carrington longitude increases from left to right for a given rotation. Curved solid lines denote the magnetic neutral lines on the source surface calculated from photospheric field observations at the Wilcox Solar Observatory with a potential field model (http://wso.stanford.edu/synsourcel.html) [Hoeksema et al., 1983]. Note that the magnetic neutral lines are determined for latitudes between −70° and +70°.
 As demonstrated in Figure 1, the solar wind changed its speed distribution systematically throughout the 11-year sunspot cycle. During the solar minima, the high-latitude to midlatitude regions were occupied with the fast wind, while the low-latitude regions were dominated by the slow wind. It is noteworthy that an excellent association between the slow wind and the magnetic neutral line is shown on the IPS speed map during 1996, i.e., the 22/23 minimum (Figure 1c). The slow wind became ubiquitous at all latitudes during the maximum, whereas the fast wind greatly diminished or disappeared. This fact is basically consistent with the solar wind evolution observed during cycle 21 [Kojima and Kakinuma, 1987, 1990; Rickett and Coles, 1991]. It should be noted here that IPS speed data indicated in Figure 1 have been deconvolved by the CAT method, whereas IPS data used in earlier studies included LOS integration effects. Thus, the present study allows us to more accurately investigate long-term variations in solar wind speed distribution.
 In this study, we calculated the fractional areas of various speed components using our IPS observations, and derived their long-term variations during 1985–2008. Before proceeding further, we must pay attention to the fact that the latitude coverage of our IPS observations was nonuniform. Figure 2 demonstrates the latitude coverage over the 24-year period. Note that the period is divided into three subperiods of 8 years in Figure 2. As shown here, high-latitude regions were sometimes poorly observed, since the LOSs used for our IPS observations were relatively sparse over the poles. The coverage of the southern polar region was particularly inferior to that for the northern region owing to the low elevation of the winter Sun in Japan. Such nonuniform observation coverage may result in a bias in the present analysis if a specific speed component has a preference for certain latitudes. Here, it is worth noting that the observation coverage at high latitudes improved significantly after 1994. Figure 2 (middle) includes data for the pre–“break point” year, i.e., 1993 and those following. This improvement was due partly to the addition of a new antenna to the multistation IPS system and partly to the optimization of the radio source distribution in our IPS observations. As shown in Figure 2, the latitude profiles of observation coverage were similar to each other, except that a shift to a less pronounced latitude dependence occurred in 1994. Therefore, we consider that latitude coverage did not cause any systematic errors in long-term solar wind structure trends. Areal coverage of our IPS observations for the 11-rotation synoptic map varied from year to year, and this variation arose from the occurrence of data gaps due to serious system failures (see Figures 1d and 1e). The average areal coverage for the 11-rotation map during 1985–2008 was 67 ± 9%. We believe that this variation did not result in a significant bias in the present analysis, partly because these interruptions occurred sporadically, and partly because the magnitude of the variations in areal coverage was small.
3.1. Fractional Area Variations in High-, Low-, and Intermediate-Speed Solar Wind
 The year-to-year variations of fractional areas for <445 (red), 445–530 (yellow), 530–615 (green), 615–700 (cyan), and >700 km/s (blue) are displayed in Figure 3. The speed areal boundaries of <445 and >700 km/s limit the slow and fast solar wind areas, respectively, and the other boundaries limit the intermediate-speed wind areas. Figures 3 (bottom) and 3 (top) show the ratios to the observed area A1 and those to the whole area A2 for an 11-rotation map, respectively. In Figure 3, IPS speed data at all latitudes are used. Cyclic increases in the fractional areas of the slow (<445 km/s) wind in the maximum and the fast (>700 km/s) wind in the minimum are clearly revealed in Figure 3. In the case of A1 (Figure 3, bottom), peak values of the slow wind fractional area occurred in 1990 and 1999 (53% and 50%, respectively), while the fast wind fractional areas in three minimum periods are 33%, 46%, and 35% for 1986, 1996, and 2008, respectively. It should be noted here that, as mentioned in section 2, the fast wind fractional areas during the solar maximum may have been underestimated. The fractional area of the intermediate-speed (445–700 km/s) wind showed no periodic variation associated with the sunspot cycle, and it appeared to be approximately constant throughout the period. The averaged values of A1 and A2 for different speed components are listed in Table 1. Fractional areas of three intermediate-speed areal boundaries for both A1 and A2 data (see Figure 3 and Table 1) exhibited consistent small excursions and subsequent small standard deviations, suggesting the constancy of the intermediate-speed wind area over the solar cycle. This also suggests that the effect of nonuniform coverage from our IPS observations is insignificant.
Table 1. Averaged Values of Fractional Areas A1 and A2 for Different Speed Components
27.6 ± 11.9
21.0 ± 6.5
16.1 ± 3.1
13.0 ± 4.2
22.3 ± 13.7
20.0 ± 9.3
15.2 ± 5.5
11.5 ± 2.7
9.1 ± 2.8
15.7 ± 9.8
Figures 4 and 5 show the year-by-year variations in fractional areas A1 and A2 for ∣latitude∣ < 10° and ∣latitude∣ > 70°, respectively. The speed areal boundaries used here are the same as the ones used in Figure 3. No significant difference between A1 and A2 was discerned in Figures 4 and 5, and this supports the conclusion mentioned above. The dominance of the slow-speed (<445 km/s) areal boundary at low latitude is clearly demonstrated in Figure 4. One exception to this is seen in the data for 2003, which indicated a marked increase of high- and intermediate-speed winds. Fast (slow) wind areas at low latitudes increased (decreased) slightly during the declining and minimum phases of each solar cycle, but those changes were much less evident than in Figure 3. The increased fast wind areas have been ascribed to the equatorward extension of polar coronal holes, which resulted in large north-south excursions of the heliospheric current sheet during the declining to minimum phases, and a similar feature had been reported from IPS observations during cycle 21 [Kojima and Kakinuma, 1987, 1990]. We also note that the low-latitude fast wind area increased significantly for the late phase of cycle 23, i.e., for 2003–2008. A marked growth in the fast wind at the equator has already been reported from STEL IPS observations for 2007 and 2008 [Tokumaru et al., 2009]. On the other hand, areas of low- to intermediate-speed wind at high latitudes showed distinct cyclic variation with sharp peaks observed at solar maxima (i.e., 1990 and 2000), while those of the high-speed winds behaved inversely (see Figure 5). In other words, a strong dominance of fast winds in the high-latitude region was revealed throughout the solar cycle, except for a short period around the maximum. If we allow for the effect of errors in the CAT analysis (see section 2), the actual variations in fast and intermediate wind areas may be more moderate than the observed ones. This result is basically the same as that observed during cycle 21 [Kojima and Kakinuma, 1987, 1990]. The marked decrease in fast wind areas at high latitudes during solar maximum has been attributed to the shrinkage or disappearance of polar coronal holes during that period [Hundhausen et al., 1981]. An important point to note is that fast wind areas observed during the declining-to-minimum period of cycle 23 (2004–2008) were systematically smaller than those during the same phase of cycle 22. A more detailed description of the comparison between the 1996 and 2008 minima is presented in section 3.2.
3.2. Comparison Between 1996 and 2008 Minima
 Focusing on the years 1996 and 2008, which correspond to the cycle 22/23 and 23/24 minima, we address the differences in solar wind speed distribution between those two minima. The latitude coverage of IPS observations for 1996 was very similar to that for 2008 (see Figure 2), leading us to consider that the effect of varying latitude coverage caused no significant bias in this comparison. If IPS data in the cycle 21/22 were compared with those in the cycle 22/23 and 23/24, an effect of the solar magnetic field polarity on the solar wind speed distribution could be argued. Unfortunately, the latitude coverage for the cycle 21/22 was significantly different from that for the cycle 22/23 and 23/24, and this prevented us from making such a comparison. We hereafter show only A1 data, since A2 data give basically the same values as A1 for these years. The fractional areas A1 of different speed areal boundaries observed in 1996 and 2008 are illustrated in Figures 6a–6c for all latitudes, ∣latitude∣ < 10°, and ∣latitude∣ > 70°, respectively. A comparison between the 1996 and 2008 data revealed the following differences:
 1. The fast (>700 km/s) wind area observed in 2008 was smaller than that in 1996 by a factor of 0.73, while the other speed areal boundaries observed in 2008 were greater than those in 1996. In particular, there was a pronounced increase in the intermediate-speed (530–615 km/s) component in 2008 (see Figure 6a).
 2. The fast wind areas observed in 2008 showed a remarkable growth at low latitudes, as compared with the 1996 data. A similar tendency was found for two intermediate-speed areal boundaries at low latitudes (530–615 and 615–700 km/s), whereas a significant reduction of low-latitude areas was found in two lower-speed areal boundaries in 2008 (<445 and 445–530 km/s; see Figure 6b). Interestingly, the increase of fast wind areas at low latitudes was overwhelmed by the fast wind area decrease at other latitudes, resulting in an overall decline of fast wind at all latitudes, as mentioned above.
 3. While a strong dominance of the fast wind is shown from high-latitude data in both 1996 and 2008, a slight drop in fast wind areas at high latitudes was observed in 2008. Other speed components exhibited an increase of the high-latitude area for 2008 (see Figure 6c).
 Thus, some significant differences in the solar wind speed distribution between 1996 and 2008 were identified from our IPS observations. These differences cannot be explained by any biasing effects in the IPS observations. As reported in our previous study [Tokumaru et al., 2009], the solar wind structure observed at the cycle 22/23 minimum was quite similar to that at the 21/22 minimum, suggesting that the differences in 2008 can be regarded as a peculiarity of the 23/24 minimum. The most important aspect characterizing the 23/24 minimum is the weak polar field (see section 3.3). This is considered as the underlying cause that produced an unusual solar wind speed distribution during the current minimum [e.g., Luhmann et al., 2009]. The relation between the solar wind speed distribution and the polar magnetic field is examined in section 3.3.
3.3. Correlation With Polar Field Strength
 The physical connection between the solar wind speed and the polar magnetic field is explained as follows. The polar magnetic field plays an important role in formation of large-scale coronal structures, particularly for corona holes where the open magnetic flux of the Sun is concentrated. If the polar field is strong, large coronal holes develop at high latitudes. As the polar field decreases, high-latitude coronal holes shrink in size. The solar wind speed is known to depend on physical properties of the coronal magnetic field. An empirical inverse relation between the solar wind speed V and the expansion rate of Sun's open magnetic flux, f, has been disclosed from earlier studies [e.g., Wang and Sheeley, 1990; Hakamada and Kojima, 1999]. The central part of large coronal holes corresponds to a region with a small value of f, i.e., the fast wind source, whereas the boundary region of coronal holes or open flux areas near the active regions are associated with a large f, i.e., the slow wind source. It has been claimed from recent studies [Kojima et al., 2004; Suzuki, 2006] that the parameter B/f, where B is the photospheric magnetic field intensity, is more important in determining the solar wind speed than the empirical relation; V ∝ 1/f. Whichever is adopted as a model for the relation between the coronal magnetic field and the solar wind, the polar field is expected to strongly control the solar wind speed distribution.
Figures 7 (bottom), 7 (middle), and 7 (top) display time variations in the fractional area A1 of different solar wind speed components for all latitudes, the polar magnetic fields derived from the Wilcox Solar Observatory [Svalgaard et al., 1978; Hoeksema, 1995], and the monthly-averaged sunspot number, respectively, during 1985–2008. The polar field and sunspot number data were obtained from http://wso.stanford.edu/Polar.html and http://www.ngdc.noaa.gov/stp/SOLAR/ftpsunspotnumber.html, respectively. In Figure 7 (middle), field strengths at the north (N) and south (S) poles are plotted by solid lines, and (N − S)/2 by a dashed line. An enhancement of the polar field strength during the minimum and a reversal of the magnetic polarity during the maximum are demonstrated here. An important point to note is that the polar field strength of this minimum is about half as large as that of the previous one. Figure 7 (bottom) clearly reveals a cyclic change in A1 with the peaks which occur alternatively for fast and slow winds. As mentioned in section 3.2, the increase in the fast wind area during the cycle 23/24 minimum is less prominent than that during the 22/23 minimum. Thus, the long-term evolution of fast and slow wind areas displayed a close association with the solar cycle change in the polar magnetic field.
 The all-latitude fractional areas are plotted in Figures 8 (left) and 8 (right) as a function of the yearly mean magnitude of the polar field strength for fast (>700 km/s) and slow (<445 km/s) winds, respectively. Open and solid symbols denote data taken during the periods 1985–1996 (cycle 22) and 1997–2008 (cycle 23), respectively, while solid squares denote the data obtained in 2007 and 2008. A striking feature revealed here is the close association of the fast or slow winds with the polar field. The fast wind data exhibited an excellent positive correlation with the polar field, while an increased scatter in the fast wind data was observed at the stronger polar fields (>0.5 G). The correlation coefficient between fast wind areas and polar fields is 0.770. Note that those areas for the weak polar field may be underestimated owing to the effect of errors in the CAT analysis (see section 2). In contrast, slow wind data were inversely correlated with the polar fields, and had a negative correlation of −0.683. Since the polar field is closely associated with large high-latitude coronal holes that are the source of the fast wind, we consider that the correlation between the fast wind and the polar fields is of primary importance, and that the slow wind–polar field correlation is secondary to it. Cycle 22 and 23 data give generally the same result, although the latter, particularly for 2007 and 2008, appear to deviate from the trend to give a larger than nominal fractional area for the fast wind and a smaller fractional area for the slow wind for the determined field strength (∼0.5 G). This deviation is of almost the same magnitude as the data scatter revealed in Figure 8, so that we cannot say for certain whether the data taken in 2007 and 2008 follow a different trend. Further discussion of this discrepancy is beyond the scope of the present study, and we need to keep monitoring evolution of the solar wind structure and its relation with the polar field during cycle 24, since some peculiarities of this cycle have already been reported from remote sensing and in situ measurements [e.g., McComas et al., 2008; Tokumaru et al., 2009].
 As mentioned in section 3, STEL IPS observations show a distinct growth of the fast- and intermediate-speed winds at low-latitude regions during the late (declining through minimum) phase of cycle 23. To confirm this finding, we examined in situ data of the solar wind speeds measured by the Advanced Composition Explorer (ACE) at the L1 orbit [McComas et al., 1998]. Occurrence rate histograms of hourly average wind speeds measured by ACE are displayed in Figures 9a–9h for 8 years from 2001 through 2008. The ACE Level 2 (verified) data (http:://www.srl.caltech.edu/ACE/ASC/) are used here. The speed width of each bin is 10 km/s, and the number of data, average and mode values of wind speeds are indicated on each plot. In situ data taken in 2001 (Figure 9a) and 2002 (Figure 9b) showed a sharp peak in the occurrence rate at ∼400 km/s, suggesting the prevalence of the slow-speed component for the ecliptic solar wind during solar maximum. While the slow component was the most prevalent throughout the period, a substantial increase in the occurrence rate toward the 2008 minimum occurred at ∼600 km/s. This fact was consistent with the IPS observations at low latitudes during the late phase of cycle 23 (see section 3). In situ speed data obtained in 2007 and 2008 clearly showed a double-peak distribution, as noted in our previous paper [Tokumaru et al., 2009]. In the case of 2008 data, the occurrence peak at ∼600 km/s was almost comparable to the one at ∼400 km/s. Such a marked enhancement of fast winds in the ecliptic is ascribed to the emergence of open field regions (i.e., coronal holes) at low latitudes, which is closely linked with the weak polar field during the cycle 23/24 minimum [Tokumaru et al., 2009; Luhmann et al., 2009].
 Some peculiar aspects of the solar wind properties during the cycle 23/24 minimum have been reported in earlier studies. In situ observations during Ulysses' third orbit have revealed that the fast wind from large polar coronal holes was slightly slower, significantly less dense, cooler, and had less momentum flux [McComas et al., 2008; Issautier et al., 2008]. Ulysses observations have also indicated that the open magnetic flux in this solar minimum decreased by a factor of 0.64 compared to the previous minimum [Smith and Balogh, 2008]. The low interplanetary magnetic fields in this minimum have been observed from ecliptic near-Earth in situ measurements [Lee et al., 2009]. A close link between the excess of high-speed winds in the ecliptic and the weak polar fields has been demonstrated using model calculations [Luhmann et al., 2009]. According to their results, the weak polar field causes prominent development of high-speed winds in the ecliptic for two reasons: First, the weaker polar fields make the low-latitude/midlatitude coronal holes larger. Second, the ecliptic more often maps deep inside the midlatitude open field sources. Thus, one important way in which the solar dynamo significantly affects the quiet solar wind has been illustrated in their studies. Similarly, the unusual features of the current solar wind mentioned above are thought to be a manifestation of long-term evolutions in solar dynamo.
 Interestingly, our IPS observations suggest that the excess of high-speed winds at low latitudes began in 2003. The in situ speed data also suggest that a considerable evolution of the ecliptic solar wind occurred between 2002 and 2003 (see Figure 9). The solar wind speed map derived from our IPS observations for 2003 is displayed in Figure 10. A close association is demonstrated here between slow winds and the magnetic neutral line (solid line) that appears greatly warped. While fast winds were preferentially distributed at high-latitude regions during the period, a marked development of equatorial fast winds was observed for CRs 2004–2005 and 2008–2009. The physical connection between the increase of equatorial fast winds observed in 2003 and the unusual features in 2007–2008 is not well understood. Further research into this connection is beyond the scope of this paper, although it may provide an important insight into the solar dynamo.
 STEL IPS observations were used to study the solar cycle variations in the solar wind speed distribution during 1985–2008. As a result, the fast (>700 km/s) wind areas on the entire source surface were found to increase as the solar activity declined, reaching its maximum in the solar minima. The slow (<445 km/s) winds were found to behave inversely to the fast wind. The areas of intermediate-speed (445–700 km/s) wind were approximately constant during the period. The low-latitude (<10°) region was mostly dominated by the slow wind over two solar cycles, while a strong prevalence of the fast wind in the high-latitude (>70°) region was revealed throughout the period, except for a few years around two solar maxima in 1990 and 2000. These facts are consistent with what had been observed in cycle 21 [Kojima and Kakinuma, 1990; Rickett and Coles, 1991]. However, distinctly different features were identified for the declining phase to the subsequent minimum of cycle 23. Those are a significant increase in the low-latitude fast wind area and a noticeable decrease in the high-latitude fast wind. These results, which are consistent with in situ measurements, are ascribed to the weak polar magnetic field in this minimum, namely a peculiarity of the solar dynamo. Our IPS observations were compared with magnetograph measurements at the Wilcox Solar Observatory, and consequently an excellent positive (negative) correlation was revealed between the fast (slow) winds and the polar fields. This fact confirms a strong control of the coronal magnetic field in determining solar wind formation.
 The poloidal field of the Sun is thought to be closely linked with the toroidal field activity in the maximum phase via the dynamo process, and thus the evolution of the Sun's magnetism over the course of cycle 24 is likely to differ significantly from other solar cycles within past the 50 years. Accordingly, the solar wind is likely to show different global features during cycle 24. Our IPS observations are useful in elucidating how the global solar wind structure evolves during these solar activity conditions that are unprecedented in the space age.
 The IPS observations were carried out under the solar wind program of the Solar-Terrestrial Environment Laboratory (STEL) of Nagoya University. This work was partly supported by a Grant-in-Aid for Scientific Research (B) (21340140) and a Grant-in-Aid for Creative Scientific Research (17GS0208) from the Ministry of Education, Culture, Sports, Science and Technology of Japan. We would like to thank the ACE Science Center and Wilcox Solar Observatory for providing open access to the solar wind plasma data collected by the ACE spacecraft and the magnetogram data, respectively.
 Amitava Bhattacharjee thanks W. A. Coles and another reviewer for their assistance in evaluating this paper.