Intraseasonal oscillations of the Silk Road pattern lead to predictability in East Asian precipitation patterns and the Mei Yu front

The Silk Road pattern (SRP) is analysed on intraseasonal timescales over summer using empirical orthogonal functions (EOFs) of the meridional wind at 200 hPa. The first two principal components explain almost equal amounts of variance, hence both are required to represent the intraseasonal SRP. The associated spatial loadings are 90° out of phase with each other, providing evidence that propagating oscillations are a natural mode of variability of the intraseasonal SRP. This is supported by Hovmöller diagrams of the meridional wind at 200 hPa and by phase diagrams of the first two EOFs, which both show a predominantly eastward-propagating oscillation. The oscillations are identified as plausibly being Rossby waves by means of waveguide theory. The subtropical westerly jet and East Asian rainfall patterns are found to be dependent on the phase of the oscillation: wet anomalies occur to the east of troughs in the jet, which are also regions where local jet entrances cause upper-level divergence via an ageostrophic circulation. Dry anomalies occur to the west of troughs, which are regions of upper-level convergence. The time-delayed location of the summer Mei Yu front relative to its climatology is dependent on the phase of the oscillation: when there is an upper-level trough located over the Korean Peninsula, the Mei Yu front is likely to be located further north than normal 3–8 days later, before returning to its climatological position. This suggests that the phase of the intraseasonal SRP acts as a potential source of predictability of the location of the Mei Yu front, which might allow for better prediction of the associated rains.


Introduction
The Silk Road pattern (SRP) is a large-scale summertime teleconnection of the upper-level winds over Europe and Asia (Enomoto et al 2003). It is a manifestation of the Rossby waves that form in the subtropical westerly jet (SWJ), where the SWJ acts as a waveguide due to the meridional change in zonal winds (Hoskins and Karoly 1981, Enomoto et al 2003, Stephan et al 2018. It can be considered as the Asian branch of the circumglobal teleconnection (Ding and Wang 2005). On seasonal timescales, it has been shown to have an effect on rainfall patterns over Asia (Wang et al 2017, Stephan et al 2018. The SRP is commonly studied using empirical orthogonal function (EOF) analysis, in which the EOFs are calculated using the meridional winds at 200 hPa, as these capture the Rossby wave motion. The EOF analysis is normally performed on monthly (e.g. Yasui and Watanabe 2010) or seasonal (e.g. Stephan et al 2019) timescales, in which the purpose is to determine the activity of the SRP over a given summer season and, for example, link this to Asian rainfall (e.g. Wang et al 2017, Stephan et al 2019. It is less commonly studied on intraseasonal timescales-in Stephan et al (2019) they performed analysis on daily-mean time periods, although the purpose of their study was to analyse the effect of interannual SRP variability on its daily variability. In Ding and Wang (2007), intraseasonal variations of the Eurasian summer wave train are analysed using daily EOF analysis, and they found that there was evidence of 'an eastward and equatorward propagation of a Rossby wave train from the northeastern Atlantic to East Asia'. However, their analysis domain is substantially larger than ours, and hence their centres of action do not align with the ones we find.
The purpose of this study is to analyse the SRP on intraseasonal timescales, by performing EOF analysis on the daily upper-level meridional winds. Following previous results at seasonal timescales, we will identify links between the intraseasonal SRP and East Asian precipitation. Furthermore, links have been made between the seasonal SRP and the location of the Mei Yu front (MYF) (Kosaka et al 2011), which is a summertime large-scale atmospheric front, oriented approximately zonally and which travels north across much of East Asia from April to August. On short timescales the location of the MYF can have a strong effect on Chinese rainfall (Volonté et al 2021(Volonté et al , 2022, which studies of the seasonal mean will not be able to detect. Therefore, we will perform a novel investigation of whether there are links on intraseasonal timescales, determining whether there is a time-delayed or lagged relationship between the intraseasonal SRP and the location of the MYF relative to its climatological location. This will provide some evidence of short-timescale predictability between the intraseasonal SRP and East Asian weather systems, and shed light on the role of some extratropical drivers on the MYF. Recent years have seen several occurrences of extremely heavy rainfall over East Asia (e.g. , Volonté et al 2021, highlighting the need to assess the predictability of such events.
In section 2 we briefly describe the datasets and EOF analysis that we employ. In section 3, we present our findings, giving evidence for a propagating phase relationship of the intraseasonal SRP and then describing its effect on East Asian meteorology. We finish with a brief summary and discussion in section 4.

Data and methods
For the upper-level winds and divergence, we use the 200 hPa hourly fields from ERA5 (Hersbach et al 2020) for the summer months of June-July-August (JJA) over the years 1979-2020, on a horizontal resolution of 0.25°. We take four hourly fields per day at 0000, 0600, 1200 and 1800 UTC and average these to a daily field. We use the Global Precipitation Measurement (GPM) Integrated Multi-satellitE Retrievals for GPM (IMERG; Huffman et al 2020) dataset for precipitation observations, for JJA over 2000-2020, using the 'Final' version, which is suitable for scientific analysis.
The EOF analysis is performed on the daily 200 hPa meridional wind field from ERA5. We perform the analysis as described in Wilks (2011) and implemented by Pedregosa et al (2011). As in Stephan et al (2019), we use a domain of 30-130°E, 20-60°N. We also tested the analysis on the seasonal JJA-mean of the meridional wind field, which produced similar results to previous work, although we do not show the results here.
To determine the location of the MYF, we use data from the front detection algorithm developed by Volonté et al (2022), which is an automated algorithm based on the meridional gradient of equivalent potential temperature at 850 hPa and is designed to detect the East Asian summer monsoon front-the summertime western portion of which is the MYF. We calculate the MYF's latitudinal location over China by averaging over longitudes of 112-122°E, and define the front as being northward or southward of its climatological position if it is over 0.75 standard deviations of its daily variability from its climatological position for the time of year. We determine the lead-lag relationship between the intraseasonal SRP and the MYF by producing a lagged contingency table (Cassou 2008, Lee et al 2019 based on the phases of the intraseasonal SRP and the location of the MYF relative to its climatology.

Propagating oscillations of the intraseasonal SRP
The EOF analysis that we employ is a statistical tool that detects the principal components that explain the most variance of the analysed data. In this section, we present evidence of a phase propagation of the intraseasonal SRP, through EOF analysis of the upper-level meridional winds. This result arises as a natural interpretation of the EOF analysis; we did not specifically set out to find phase propagations. Figure 1 shows a scree plot (Wilks 2011) for the first 10 daily EOFs. It is clear that EOF1 and EOF2 explain much more variance than the remaining EOFs and notable that they explain almost identical amounts of variance. This is reminiscent of the findings of Wheeler and Hendon (2004), in which they developed a phase diagram for the MJO-i.e. there are a coupled pair of EOFs that are 'degenerate' and therefore cannot be used in isolation. We note that in Wheeler and Hendon (2004) such a pair of EOFs with almost equal explained variances indicated that the EOFs should be well suited for describing a travelling wave, and this also applies here. For seasonal EOFs (not shown), there is far greater variance explained by the first EOF (35%) than the second (15%), which indicates that only the first EOF is needed on a seasonal timescale. When analysing a larger domain that encompasses ours, Ding and Wang (2007) found that the first three EOFs are degenerate, although the difference in domains means that their centres of action are displaced from the canonical SRP centres of action. Furthermore, they found that their EOFs 2 and 3 formed an intraseasonal oscillation, which they associated with a propagating Rossby wave train. Figure 2 shows the spatial loadings of daily EOF1 and EOF2, along with their values through a transect running approximately through their centres of action, where all values have been scaled by the standard deviation of their principal components (PCs). From figure 2(c), it is notable that the two EOFs are approximately in quadrature (i.e., 1 4 of a wavelength, λ, or 90°out of phase with each other). This indicates that the first two EOFs would be well suited to representing waves of any phase or propagating waves in the upperlevel meridional winds. Daily EOF1 loadings are very similar to the same for seasonal EOF1 (Stephan et al 2019) (note, the sign of the spatial loadings is arbitrary). EOF2 bears a strong similarity to seasonal EOF2 as well (not shown). From figure 2(c), we estimate the wavelength of EOF1 to be 52°of longitude at 36°N, or λ x = 4700 km. EOF2 has a very similar wavelength. Figure 3 shows Hovmöller plots for four selected years, as well as the overlayed projections of EOF 1 and 2. The years were chosen to emphasize certain features: 1980 has clear signs of an eastward-propagating oscillation from mid to late June; 1981 shows some oscillations travelling westward. 1984 and 2016are the years with strongest and weakest seasonal SRP respectively (see Stephan et al (2019) for 1984; 2016 not shown). We note that 2016 was a year with heavy Mei Yu rainfall, although the possible connection between this and the weaker seasonal SRP is not explored further here.
1980 shows clear signs of eastward propagation, particularly towards the end of June. This is detected by daily EOFs 1 and 2. Even when the EOF amplitudes are small (e.g. in August), there are signs of propagating oscillations. The late-June oscillation has a phase speed of approximately 5°longitude per day (at 40°N) or approximately 5 m.s −1 (estimated by fitting lines to the Hovmöller diagrams). This is consistent with Hunt et al (2018), in which they found a phase speed propagation of 4.7 m.s −1 along the SWJ in winter. 1981 shows some signs of westward propagation, at the start of July, which is picked up by the EOF analysis. This appears to be a wave packet, and the EOFs are detecting its group velocity; eastward propagating features are apparent at the same time (e.g. 40-50°E, at the start of June).
1984 has a strong positive phase of EOF1 (compare figures 2(a) with 3(c)-the spatial pattern of EOF1 is dominant) from 20 July onwards, and sporadic strong phases of EOF1 before this. There are weaker signs of propagating oscillations, e.g. around mid-July. 2016 has strong negative phase of EOF1 (see figure 2) from mid-August onwards, and some strong negative phases of EOF1 before this. Interspersed with these are clear signs of propagating oscillations. These examples illustrate that there is considerable variability in the intraseasonal SRP, and that eastward and westward propagation, as well as being stationary, are possible. Analysis of all other years (not shown) shows that eastward propagation appears to be dominant.   Figure 4(a) shows a phase diagram of PCs 1 and 2, in a similar manner to Wheeler and Hendon (2004). We have labelled the phases according to their location in figure 4(a): magnitude <1 is phase 0; phases 1-4 are defined where magnitude >1 and by quadrants where 1 is positive PC1, 2 is negative PC2 etc. The reason we chose to label 2 as negative PC2 is because we wished the phases 1, 2, 3, 4 to appear in succession with the dominant propagation direction, and a clockwise rotation represents an eastward propagation. The year 1980 is highlighted (also shown in figure 3); in JJA it shows a clear clockwise (eastward) propagation. Figure 4(b) shows the distribution of angular rotation between each day-i.e., if the progression on the phase diagram is −12°(clockwise) for one day, it will contribute to the −10 to −20°bin in the density plot. Figure 4(b) shows that negative (i.e. clockwise, representing eastward propagation) rotations are more likely than positive, for both days where there is a strong combined EOF 1 and 2 and for all days. For both distributions, the mean is significantly different from zero, using a two-sided t-test with a 5% rejection level. For days where magnitude >1, there are approximately three times as many days where the propagation is eastward compared to westward (1490 compared to 425 days). In both cases, the modal angle is −10 to −20°(N.B. this is angle of rotation of phase per day, not longitude; conversion to degrees longitude per day or phase speed can be performed by taking the wavelength in degrees longitude or metres respectively). This corresponds to a phase speed of 1.5 m.s −1 to 3.0 m.s −1 eastward. Phase speeds of propagating waves travelling between 4.5 m.s −1 and 6 m.s −1 , i.e. including 5 m.s −1 as found above in the analysis of figure 3 for the propagating wave, occur 8.7% of the time when magnitude >1 and 8.2% for when magnitude <1.
Previous studies have shown that the SWJ can support Rossby waves (Enomoto et al 2003, Hunt et al 2018, and this has been postulated as an active process on the SWJ over summertime (Ding and Wang 2007, Li and Lu 2017, Wang et al 2017. We therefore performed analysis of the expected phase speed of the wave disturbance, assuming the wave is a Rossby wave and that the SWJ acts as a waveguide Karoly 1981, Enomoto et al 2003). From wave theory of the linearized baratropic vorticity in the presence of a waveguide, a Rossby wave's phase velocity is given by = -b + c U p k l eff 2 2 , where β eff = β − U yy , and U yy is the second meridional derivative of the zonal wind (Enomoto et al 2003). We calculated U by taking the value of the mean zonal wind at 200hPa over the years 1979-2020 from ERA5 in the domain bounded by the contour of U = 20 m.s −1 where U > 20 m.s −1 , and likewise for U yy . We estimate k from the zonal wavelength of EOFs 1 and 2: λ x = 4677 km, giving k = 1.34 × 10 −6 m −1 , and we take l = 0 for assumed no meridional propagation. β is calculated at 36. 5°N . This gives a theoretical phase velocity of 0.12 m.s −1 , which is lower than the observed value here for the detected propagating wave. However, it is close to the value derived from the mode of the rotation angles in figure 4, and is well within the uncertainty defined by the half-width of the PDF. Thus, this provides limited evidence that the daily EOF analysis could be detecting the phase of propagating Rossby waves formed in the SWJ waveguide.

SWJ and East Asian precipitation conditional on phase of the intraseasonal SRP
We have identified that the intraseasonal SRP can have different phases. Here, we ascertain the effect of these phases on East Asian precipitation patterns. Figure 5 shows precipitation and upper-level winds conditional on the phase of the daily EOFs 1 and 2. Phase 0, corresponding to when there is little intraseasonal SRP activity, is close to the mean over the given period from 2000-2020 (not shown), and shows prominent features such as heavy rainfall over the coast of Myanmar, the Western Ghats in India and the Himalayas, associated with the Indian summer monsoon. Associated with the East Asian summer monsoon, large areas of rainfall are present over China, although the rainfall is lighter than the heavier rainfall caused by the Indian summer monsoon.
For the SWJ, a trough progressing eastward is clear during phases 1-4, centred over approximately 90°E, 110°E, 120°E and 130°E respectively. These longitudes correspond to the locations of the EOF spatial loadings in figure 2. The trough is associated with a region of upper-level convergence upstream (west) and divergence downstream (east). These are likely due to ageostrophic effects at the jet exit and entrance regions respectively. Upper-level convergence is associated with subsidence. Correspondingly, these are regions of negative rainfall anomaly (e.g. phase 4 over China and the Yellow Sea). Conversely, regions of divergence are associated with Figure 5. East Asian precipitation, SWJ and upper-level divergence conditional on the phase of daily EOFs 1 and 2. Left column shows how each composite precipitation field was calculated: (a) was the mean over all points in phase 0 (i.e. magnitude < 1), and (b)-(e) show the difference between the phase in question (shown by red in left column) and phase 0 (shown by blue). For (b)-(e), coloured contours show precipitation differences, with statistically significant differences shown by stippling using a two-sided Wilcoxon-Mann-Whitney rank-sum test with a 2% rejection level (Wilks 2011, pp. 159-161). Purple lines show the position of the SWJ, using 200 hPa wind speeds contoured at 20 and 30 m.s −1 . Green solid and dashed lines show regions of upper-level (200 hPa) divergence and convergence respectively, with values of absolute magnitude exceeding 1 × 10 −6 s −1 shown. The divergence field was smoothed using averaging over 10 × 10 grid boxes (with a horizontal grid length of 0.25°) to remove some of its noise. upward motion, which is co-located with areas of heavier rainfall (e.g. phase 3 over the Yellow Sea and the Korean Peninsula).
Each phase shows a distinct pattern of rainfall from phase 0 (significant at a 2% level, using a Wilcoxon-Mann-Whitney test, unless indicated). Some notable features are that phase 1 has more precipitation over large regions of inland China, including the Yangtze river basin (YRB). Li and Lu (2017) also found connections between YRB rainfall anomaly and the meridional anomalies of the jet, however their analysis was performed on monthly data and so they were investigating the same phenomenon on a different timescale. Phase 2 has higher rainfall over the western end of the Himalayas, and the Yellow Sea and Korean Peninsula. Phase 3 has again higher rainfall over the Yellow Sea and the Korean Peninsula, as well as over the Sea of Japan, although these are not statistically significant. Phase 3 also has lower rainfall over inland China. Phase 4 has a large dry anomaly over China, and a very large and strong wet anomaly over the Philippine Sea. Phase 4 shows increased rainfall over the Tibetan Plateau. This is consistent with Wang et al (2018), who show a filtered upper-level geopotential height anomaly which is equivalent to our phase 4 (compare their figure 2(a) and figure 2(a) above). Thus, the phase of the intraseasonal SRP is a driver of East Asian rainfall when analysed over multiple years, and a plausible mechanism-the upper-level convergence and divergence driving large-scale descent and ascent respectivelylinking the two has been established.
To further explore the mechanism, figure 6 shows vertical cross sections over the region roughly aligned with the trough in figure 5(d) (phase 3) averaged over 120-130°E. This region encompasses the Yellow Sea (33-40°N ), which we will focus on here. The dry and wet anomalies in the Yellow Sea seen in figure 5 are clearly visible. The dry anomalies can be seen in phase 1 between 28-39°N and phase 4 between 34-43°N. The wet anomalies are evident in phase 2 between 35-45°N and phase 3 between 33-43°N. The maximum dry anomaly is approximately a factor of 2 smaller, and the maximum wet anomaly is approximately 50% larger than the precipitation in phase 0. The dry anomaly is associated with weaker (phase 1) or even negative (phase 4) vertical motion, and the wet anomaly is associated with stronger than phase 0 vertical motion. Both dry anomalies are associated with upper-level convergence, consistent with our hypothesized mechanism, although this is weak for phase 1. For both wet anomalies, there is substantial divergence aloft.
The position of the SWJ varies across the different phases, although the stronger 20 m.s −1 contour does not show a great degree of movement on its southern edge. The SWJ maxima in phases 1 and 2 are 5°further north than for phases 3 and 4. The northern edges of both contours shows a clear progression through the phases, with maximum extent at phase 1 and minimum at phase 3, consistent with figure 5. The location of the downward vertical winds and upper level convergence is clearest in phase 4, and is consistent with the trough being located further east than the cross-section domain and our proposed mechanism.
Further south of the Yellow Sea, phase 4 shows a very strong wet anomaly from 20-31°N, with a maximum anomaly being approximately twice as large as the phase 0 precipitation. This is associated with stronger than phase 0 (or any other phase) updraughts at around 500 hPa for this region. The circulation appears to be associated with an overturning circulation that extends further south than 20°N, and so outside our analysis domain. This could occur because the intertropical convergence zone (ITCZ) extends further northwards in this phase, leading to ascent in our domain and divergence aloft.

Lagged location of MYF conditional on phase of the intraseasonal SRP
Given the evidence for a propagating intraseasonal SRP phase and its effect on East Asian precipitation, we wish to see whether we can link the intraseasonal SRP phase to the MYF, a dominant mode of East Asian summer atmospheric variability. Additional motivation for this analysis is provided by the result of Volonté et al (2022), who showed that the location of the MYF was influenced by the SWJ on a seasonal scale, and hence investigating their interaction on shorter timescales might help to shed light on the mechanisms of this. We consider the location of the MYF relative to its climatology at a lagged number of days after a particular intraseasonal SRP phase. Figure 7 shows a lagged contingency plot of the location of the MYF relative to its climatological location for each phase of the intraseasonal SRP, as in Cassou (2008) and Lee et al (2019). This is calculated by using an index of each phase of the intraseasonal SRP, the leading variable, and comparing this with the index of the MYF location for different lags. As described in section 2, we define three possibilities for the location of the MYF-northward, southward and close to climatology -only the first two are shown. Two statistical significance tests must be met for the bars to be shaded: a χ 2 test on the distributions being significantly different at the 99% significance level, and a Gaussian normal test on the value being significantly different from 0% at a 95% significance level (Cassou 2008).
Two intraseasonal SRP phases stand out as having strong, statistically significant effects on the northward relative location of the MYF -2 and 4. 3-10 days after the occurrence of phase 2, the front is significantly less likely to be northward of its climatological location (left column). In phase 4, from days 3-8, the front is significantly more likely to be northward of its climatological location (left column). Note, it is possible for an individual lag to be more likely for both northward and southward-it implies that it is less likely to be close to its climatological location (not shown). In both cases, the MYF returns to its climatological location after 8 or 10 days.
Two possibilities suggest themselves for the approximately 3-day delay. The first is that, because of the eastward propagating nature of the some intraseasonal SRP waves, e.g. 3 days after phase 2, its trough and upperlevel divergence wind patterns are affecting the location of the MYF. For example, it could be that the location of the upper-level convergence (figure 5) is driving subsidence that causes lower-level divergence, which in turn means the MYF is less likely to be relatively further north. The second is that, because the SWJ is further north than the MYF, any effects (such as the one just mentioned) take time to influence the front. It could also be a combination of the above; further investigation on this point is needed in a future study.

Summary and discussion
Daily EOF analysis of the meridional winds at 200 hPa taken from ERA5 was performed, and we linked the phase of the propagating features to East Asian precipitation patterns and the location of the Mei Yu front (MYF). From the EOF analysis, we presented compelling evidence for the intraseasonal SRP being a natural mode of variability of the SWJ. The daily EOF analysis naturally determined that two EOFs were needed to describe the upper-level meridional winds, indicating that the phase of the oscillation can take any value and all phases are equivalently likely, as would be caused by a propagating wave. Daily EOF1 bears a strong resemblance to seasonal EOF1 from previous studies. Propagating waves were also found to be common. The modal direction of propagation is eastward, and the phase speed of propagating oscillations is approximately 5 m.s −1 eastward. Using wave analysis based on linearized baratropic vorticity and SWJ waveguide theory, we presented limited evidence that the waves are possibly Rossby waves.
The intraseasonal SRP has an effect on the upper-level divergence, through local jet entrance and exit regions, and ageostrophic dynamics. It consequently has an effect on East Asian precipitation patterns through induced subsidence and upward motion of air. Some of the prominent precipitation signals are: wet anomalies over inland China for phase 1; wet anomalies over the western Himalayas, Yellow Sea and Korean Peninsula for phase 2; wet anomalies over the Yellow Sea, Korean Peninsula and Sea of Japan for phase 3; and strong wet anomalies over Philippine Sea for phase 4. Thus, the phase of the wave is seen to be important for East Asian precipitation patterns. With a focus on the Yellow Sea, the wet and dry anomalies over this region were found to be closely related to the phase of the intraseasonal SRP, and the associated circulation patterns and upper-level convergence and divergence.
We found some signs of predictability: the lagged relationship between the intraseasonal SRP phase and relative location of MYF indicate that the MYF is partially controlled by extratropical drivers. In particular, phase 4 leads relative northward location of the MYF by 3-8 days. Furthermore, the propagation of waves indicates that there may be another source of predictability; if a propagating wave is detected, and there is a known response to a later phase of the wave, then a few days later this response may be predicted, although this was not explored further here.  Cassou (2008) and Lee et al (2019). Green and yellow shading shows significant positive and negative relationships respectively, where significance is determined by applying a χ 2 test on the distribution and a Gaussian normal test on the value difference from 0% occurrence. Dashed lines are shown to aid visualization but do not denote significance.
Our work here shows that the phase of waves on the SWJ is important for East Asian meteorology. This differs from previous work, in which the phase is assumed to be stationary or quasi-stationary (Enomoto et al 2003, Stephan et al 2019, and the important aspect was taken as being the group velocity of the waves in those studies. Undoubtedly the group velocity is important, as this determines the energy transport by the wave, but here we show that so too is its phase, and that daily EOF analysis provides a simple means of diagnosing this wave activity. Furthermore, it is interesting to note that the first two EOFs of the intraseasonal SRP correspond to the first two EOFs of the seasonal SRP, but that the explained variance of these is the same for the intraseasonal SRP and not for the seasonal. This indicates that there is a mechanism that acts to anchor the intraseasonal SRP in the location where it appears on seasonal timescales, such as perhaps anchoring due to orography or anchoring arising due to the location of whatever is generating variability in the intraseasonal SRP. Such a mechanism would be a good candidate for future study. One way of investigating this could be to perform low-pass filtering on the intraseasonal SRP with different thresholds for the cut-off frequency, to see how the intraseasonal and seasonal SRPs are related. We calculated the theoretical phase speed of the waves above from the mean fields describing the SWJ. It would be possible to perform a similar calculation on the daily fields, and relate this to the observed phase change of the intraseasonal SRP. The hypothesis would be that, as the SWJ strengthened, the phase speed of the waves would increase, and so one would expect to see a propagating wave under these circumstances. We made some attempt to perform this analysis (not shown), but the results were inconclusive and showed little correlation between the theoretical and observed phase speed. Refinement of this technique could provide the basis of interesting future work and lead to another source of predictability of East Asian precipitation.
Here, we analysed waves on the SWJ without determining what excited them. It would be of great interest to see what effect large-scale drivers of East Asian weather had on the intraseasonal SRP; drivers that have been shown to affect the seasonal SRP include Indian summer monsoon (Beverley et al 2021) and the North Atlantic oscillation (Liu et al 2020). This could further increase the predictability window of East Asian meteorology, and the intraseasonal SRP as introduced here provides a metric for evaluating medium-range to seasonal forecasts.
Finally, our work here has been based on reanalysis and observations. Evaluating the intraseasonal SRP in atmospheric models could determine whether they realistically represent the relevant processes, and if they correctly reproduce the effects on East Asian meteorology. High-resolution simulations struggle to properly represent East Asian precipitation (Muetzelfeldt et al 2020), and this could be, in part, due to their failure to correctly represent the large-scale drivers, such as the phase of the intraseasonal SRP.