1) Rainfall

During the years of 1999 and 2000 negative anomalies in sea surface temperatures existed in the eastern equatorial Pacific Ocean. That is, global circulations were under the influence of the cold or La Niña phase of the ENSO cycle. Based on the aforementioned SOI relationship identified by Shrestha (2000), Nepal would be expected to receive normal to above normal rainfall during these monsoons, and in fact received normal rainfall during both years according to the Nepal Department of Hydrology and Meteorology. During both monsoon seasons neighboring India received normal rainfall as a whole, although the Gujarat region of India experienced drought conditions during both years.

Rainfall totals for 1999 and 2000 JJAS (1 June to 30 September) for each network station are presented in Table 1. Aside from the very large amounts of rain that fell in this region, the most prominent feature is the strong variability. For stations with complete records in 1999, the range was 118–376 cm, a very large difference considering that this difference occurred over a distance of about 25 km (Ganpokhara–Temang). The general pattern in 1999 showed rain totals maximized at higher elevation stations along the ridges.

The rainfall patterns showed some differences in 2000. The biggest change was the near 100% increase in rainfall at Khudi, from 227 cm in 1999 to 411 cm in 2000. Such a large change raises the question of instrument reliability. Checks against manual measurements at the same location, as well as correlation with nearby gauges, indicated that the instrument was operating properly during both years. The cause of this increase is not known—the increase was not associated with any single event, but rather occurred over many events during the monsoon—but could be related to interannual variability manifesting itself at small spatial scales. Rainfall gradients were often stronger in 2000 than in 1999, depending on the stations considered.

Table 1 also shows 2-day rainfall totals for 12–13 June 1999 and 7–8 June 2000. These 2 days were, for most stations, the heaviest rainfall days of JJAS. Moreover, for most stations in either year, the rainfall during these time periods met or exceeded the total rainfall for the prior month of May. These rain events followed about 1 week of little to no rain at the Marsyandi network, and marked the beginning of a period of daily or near-daily rainfall lasting well into September of either year. Therefore, we define these rain events as marking the onsets of the monsoon in central Nepal.

For 8 of 20 stations these 2 days accounted for over 10% of the total rainfall, in one or both seasons. These stations tended to include the highest-altitude ridge stations (Rambrong, Sundar, Danfedanda), dry stations to the north of the Annapurnas (Temang, Monastery, Nargaon), and low-elevation stations further out on the plains to the south of the core network (Tansen, Bortung). In fact, some of the northern dry stations received more than one-third of their total monsoon rainfall during the 2000 onset.

In general, the relative spatial distribution of rainfall was similar during both onsets. Rainfall increased with elevation, with some deviations, and the highest elevation station, Rambrong, recorded the most rainfall. The pattern of very large rainfall gradients at the seasonal scale also translated to these shorter time periods associated with the monsoon onsets. As an example of this, during 12–13 June 1999 the Rambrong and Syange stations, separated by only 16.2 km, differ in 2–day rainfall totals by 280 mm, an amount comparable to the annual rainfall in the Front Range region of Colorado. The difference was nearly 400 mm in 2000. Despite the similar spatial distributions between the two onsets, the rainfall accumulation during the 2000 monsoon onset at Rambrong was much greater than in 1999, while other gauges generally received 50%–100% more rain during the 1999 onset than in 2000.

Figure 2 shows time series of rainfall for several network gauges for the 1999 (Fig. 2a) and 2000 (Fig. 2b) onsets. The 1999 onset had a strong signal at every gauge, which was consistent with a single, large, long-lived precipitation system. Interestingly, the highest-altitude station, Rambrong (4435 m MSL), dominated many other stations in terms of rainfall intensity, especially early. However, for all gauges 30-min rain rates generally stayed below 30 mm h⁻¹, indicating only light to moderate rain rates even at peak.

The 2000 onset was a longer episode than 1999, covering approximately three days (two heavy days along with a weaker third). Nevertheless, the rain rates tended to be lower than in 1999 except at Rambrong. Unlike the 1999 onset, the main event in 2000 consisted of basically two parts, with a short lull between them. Consistent with its high rainfall total during the 2000 onset (46 cm), Rambrong dominated the other gauges in both intensity and duration of rainfall, although rain rates were similar to 1999. Thus, during both onsets it was duration, rather than intensity, which was responsible for the high rainfall totals. Also, during both onsets intense rain events showed no preferential timing, as opposed to the seasonal average of a nocturnal maximum in rainfall for this region as reported by Barros et al. (2000). This suggests that local mesoscale processes were dominated by large-scale weather.

The relative importance of convective versus stratiform precipitation during these onsets was assessed with the objective of characterizing the morphology of the rain systems. Generally, it is accepted that the intensity of stratiform rain seldom exceeds 10 mm hr⁻¹ (e.g., Steiner et al. 1995). But convective rain is not limited to a 10 mm hr⁻¹ lower bound. Thus, while we can identify with good confidence convective rainfall when the rain rate is equal to or exceeds 10 mm hr⁻¹, it is much more difficult to classify rain rates below this threshold. In this work, we used both the time series of tip counts at individual gauges and 30-min rainfall accumulations to categorize rainfall during each 30-min period.

First, we calculated the percentage of total rainfall corresponding to long-lived convection (i.e., convection that lasted long enough to produce a significant average rain rate over 30 min) by comparing the rainfall from times when gauge rain rates met or exceeded 10 mm hr⁻¹ to the total rainfall. In both years, stations that received the most onset rain (Rambrong, Bortung, Tansen) had by far the largest percentage (>60%) of rainfall from long-lived convection. A significant reduction in percentage of long-lived convective rainfall occurred in 2000 versus 1999; the average among all stations dropped from 46% in 1999 to 19% in 2000. This suggests that long-lived convective processes were more important for total rainfall during the 1999 onset than in 2000.

Tip data (from the 11 gauges that recorded them) were used to identify 30-min periods that contained short-lived convective rainfall (i.e., convection that did not last long enough to produce a significant average rain rate over 30 min, but that did exceed 10 mm hr⁻¹ for a portion of that time). Almost all stations saw at least some short-lived convective rainfall, usually amounting to 25% or less of rainfall totals.

The remaining fraction of unclassified rainfall presumably was either light convective (R < 10 mm hr⁻¹) or stratiform rain. This fraction was highest for low-elevation stations that received the least rain during the onsets, as well as for dry stations north of the Annapurnas. It is reasonable to expect that stratiform rain was a significant part of this remaining fraction of rainfall, which for many stations approached or exceeded 50%, and that stratiform rain was present at most gauges during the 1999 and 2000 onsets. Thus, according to the gauges, a mixture of convective and stratiform rain likely was present during the onsets. Convective rain was most important at the gauges that received the most rainfall, whereas stratiform or light convective rain was most important at the gauges that received the least rainfall.

Precipitation structure was investigated further with TRMM–Precipitation Radar (PR) observations of the 1999 and 2000 onsets. Rain rates are well known to vary substantially on very short temporal (<5 min) and spatial scales, particularly in convective situations. Thus, instantaneous measurements of the 4-km TRMM footprint by TRMM will not necessarily represent the 30-min measurements at the rain gauges. Although the PR products underestimate rainfall rates in this region (e.g., Barros et al. 2000), they provide the only means to examine the vertical structure of precipitating systems and storm morphology in this region of the world.

The TRMM satellite passed over the Marsyandi network around 2220 UTC on 12 June 1999. This was immediately after the main peak in rainfall at most gauges. Figure 3a illustrates PR horizontal and vertical cross sections of rain estimates in the Marsyandi region. The data show a number of convective cores in the region, but at this time the heaviest rains were not occurring at network stations, consistent with the fact that this was past the time of peak gauge rainfall. The convective cores were embedded within a large region of light rain (<5 mm hr⁻¹). The vertical profile of PR rain data shows what appeared to be shallow stratiform rain at the network, which was almost completely located within the light rain region at this time.

In 2000, the TRMM satellite passed over the Marsyandi network around 2230 UTC on 6 June. Figure 3b shows PR rain estimates for this time, which was after a major peak in rainfall at Rambrong. Again, heavy rains were not occurring at the network, consistent with the low observed gauge measurements around this time (generally under 10 mm h⁻¹). The region was being affected by light rain, with some embedded convection to the south. Like 1999, the vertical cross section depicts shallow stratiform rain at the network.

Data from the Visible and Infrared Scanner and the TRMM Microwave Imager aboard the TRMM satellite were also examined. The VIRS observations and the TMI brightness temperature fields support the PR observations during both onsets, and were consistent with the presence of deep convection in the area, but not at the network gauges during the overpasses. Overall, the TRMM data reveal that the onsets storms were characterized by a mixture of convective and stratiform rain, mainly in the form of embedded convection within a larger area of light rain, which encompassed much of the Marsyandi region. This inference is consistent with the gauge data.

2) Winds

Figure 4 shows 850–500-hPa averaged U and V winds, along with average relative humidity in the same layer, as functions of time at a grid point (27.5°N, 85.0°E) close to the Marsyandi network, for both the 1999 and 2000 onsets. An increase in easterly wind strength concurrent with increases in southerly flow and relative humidity occurs in both years around the time of the onset of continuous rain at the network (∼0600 UTC on 11 June 1999, after 0600 UTC on 6 June 2000).

Comparing Figs. 2 and 4, southerly (upslope) flow in the 850–500-hPa layer continued throughout the rain event in the vicinity of the network, and the termination of rain and moist (RH > 70%) upslope flow were largely coincident. The role of moist, synoptic-scale upslope flow in enhancing convection due to an influx of low-level moisture and lifting of conditionally unstable air has been noted in other parts of the world, such as the Front Range of Colorado (e.g., Doswell 1980). This mechanism is consistent with our analysis along the south slopes of the Himalayas.

3) Thermodynamic environment

There are four radiosonde stations in India that are relatively close to the network. From these stations, the soundings immediately prior to the onset of sustained rains at the Marsyandi network were examined for each onset, as these soundings represent the best chance to sample the prestorm environment. (This worked out to be 1200 UTC on 11 June 1999 and 1200 UTC on 6 June 2000.) In 1999, the average CAPE for surface mixed layer (i.e., lowest 50 hPa) parcels was 439 J kg⁻¹. (Parcels taken from higher altitudes—925–850 hPa—contained lower CAPE values.) Average precipitable water was estimated to be 56.2 mm. In 2000, the average CAPE was 1484 J kg⁻¹, and the average precipitable water was 59.4 mm.

Despite the low CAPE observed in 1999, the sounding data from both onsets are still consistent with the prodigious rainfall observed. In 1999, all of the soundings from the 1999 onset contained potentially unstable layers, so any large-scale lifting should have led to further destabilization as the air mass approached the network region and those layers were lifted and cooled (e.g., Barros and Kuligowski 1998). Also, during both years CAPE values either grew to (1999) or were sustained at (2000) more than 1000 J kg⁻¹ during subsequent days.

1) Synoptic overview

Figures 5a and 5b show a sequence of wind streamlines and speeds at 850 hPa over the expanse of south Asia during the time period associated respectively with the 1999 and 2000 monsoon onsets in central Nepal. During both onsets, a low pressure system (hereafter termed monsoon depression) developed in the Bay of Bengal, and then proceeded northwestward toward the Himalayas. The 2000 depression formed southwest of the 1999 genesis location, and that resulted in its track being to the west of the 1999 depression. The arrival of both depressions at the mountains coincided with the onset of moist upslope flow and the beginning of rain at the network, as the southerly to southeasterly flow on their northeastern flanks impacted the Himalayas.

This effect is elucidated in Figs. 6a–b, which show longitude–height cross sections (at 27.5°N) of meridional wind and relative humidity at 0000 UTC on 12 June 1999 (Fig. 6a) as well as at 0000 UTC on 7 June 2000 (Fig. 6b). The eastern flank of the monsoon depressions created southerly midlevel jet structures in the eastern third of the domain as the depressions encroached upon the mountains. The network region resides in the lower quadrant of this jet. For either year, the jet was responsible for substantial moisture flux toward the Marsyandi network, and its evolution and decay mirrored the onset and termination of rain at the network.

2) Relationship to the Asian monsoon

ECMWF analyses for May and June of both years revealed that these depressions followed the development of the monsoon onset vortex in the Arabian Sea, and occurred when the Somali jet was well developed. These two events typically precede the onset of monsoon rains over SW India (Krishnamurti et al. 1981). In 1999, a vortex appeared in the Arabian Sea around mid-May, followed in late May by a weak low pressure system in the Bay of Bengal, which propagated northeastward over Bangladesh. The next low pressure system was the depression responsible for the onset of the rainy season in Nepal. In 2000 there was an Arabian Sea vortex that formed in late May. The subsequent depression formed in the Bay of Bengal was responsible for the 2000 monsoon onset in Nepal. In both years the depressions were associated with increased low-level westerly flow over much of the Indian subcontinent (see Fig. 5), suggesting that the Nepal onset occurred as the Indian monsoon was developing and spreading northward. Indeed, the paths of the 1999 and 2000 depressions brought them over northern India, linking the onsets in central Nepal and northern India by the passage of these storms.

This link between the two regions is demonstrated by Meteosat-5 imagery. Figure 7 shows a sequence of Meteosat-5 IR images for the 1999 (Fig. 7a) and 2000 (Fig. 7b) onsets. For 1999, the imagery shows convective activity associated with the trajectory of the depression after leaving the Bay of Bengal, then crossing northern India, and finally crashing into the Himalayas in the vicinity of the network. At first, widespread convection associated with this system existed over much of northern India. However, as the storm approached the mountains its convection became more organized and localized, and rotation was more apparent. (Note the classic spiral arm structures overrunning Nepal at 0600 UTC on 12 June.) However, the convective organization clearly decayed as the system collided with the mountains (2300 UTC on 12 June—roughly coinciding with the 1999 TRMM overpass, Fig. 3a), and convection in the region took on a more disorganized appearance.

For 2000, the imagery shows somewhat disorganized convection associated with the monsoon depression approaching the mountains from the southeast, again crossing northern India, although its track was situated west of the 1999 track. Like 1999, as the storm approached the mountains the convection appeared to organize and spin up. By 2300 UTC on 6 June 2000, there is a distinctive vortex structure, complete with spiral arms. One of the spiral arms extended to the middle Himalayan region in the vicinity of the network (nearly coincident with the 2000 TRMM overpass, Fig. 3b), although the main convection did not impact the Marsyandi region as in 1999. The organized vortex structure did not last long, as the system quickly devolved and spawned widespread convective clouds across northern India and Nepal.

The increasing convective organization and rotation as the systems approached the mountains calls for further analysis. Overlays of satellite and ECMWF data revealed that this region of organized convection actually existed in the northeastern quadrant of the main depression, in the southerly to southeasterly flow (Figs. 8a–b). This is contrary to the climatology of monsoon depression structure with convection predominant in the SW quadrant (e.g., Das 1987). Both of these depressions exhibited a similar structure earlier in their lifetimes, with convection focused near their core and SW quadrant. It was only at later times, as the depressions encroached upon the Himalayas, that the convective activity shifted to the northeastern flank and became organized. This phenomenon is examined in more detail in the next section.

The ECMWF analyses and Meteosat data suggest an explanation for the 33%–50% decrease in rainfall from 1999 to 2000, as well as for the increased convective rainfall fraction observed in 1999, which relates to the trajectories of the monsoon depressions after landfall. The 2000 depression's track was situated west of the 1999 track. Consequently, the main convection on the northeastern flank of the depression impacted the network area in 1999 (visible in the TRMM-PR image, Fig. 3a), but only grazed this area with rainbands on its periphery in 2000 (one of which is visible in the TRMM-PR image, Fig. 3b). The tracks of the monsoon depressions critically affected how much rain fell during both onsets, and they are therefore a critical element of the interannual variability of rainfall in central Nepal.

3) Storm tracks

According to the climatology, monsoon depressions from the Bay of Bengal are known to travel in the west-northwest direction along the seasonal monsoon trough that resides over northern India (e.g., Das 1987). The depressions associated with the central Nepal onsets in 1999 and 2000 moved more northwestward than to the W or WNW, and demonstrated substantial convective organization even as they arrived near the Himalayas. Das (1987) notes that monsoon depressions sometimes move northward, especially late in their life cycle. In addition, Potty et al. (2000) reported on a modeling study of four monsoon depressions, one of which traveled almost due north over central India. ECMWF analyses from the years 1997–2001 suggest that northwestward propagation is not at all unusual, at any point in the monsoon season. Thus, the motion of the 1999 and 2000 depressions is not unique, but their motion along with their increasing convective organization are very interesting.

When the 2000 depression formed, the eastern edge of the surface monsoon trough that normally resides over northern India was oriented slightly more to the south than the monsoon trough at the time of the 1999 depression. As is typical with monsoon depressions (e.g., Das 1987), both systems formed on the eastern edge of this trough, so this accounts for the difference in depression genesis locations between the two onsets. Both depressions moved within the monsoon trough.

Many studies examining the motion of monsoon depressions have used the concept of vorticity balance (Rao and Rajamani 1970; Krishnamurti et al. 1976; Douglas 1992). The general conclusion has been that these systems follow low-level vorticity generation by convergence on their western flank. Using the ECMWF data, each component of the vorticity equation (horizontal and vertical advection, divergence, and tilting) was calculated during the 1999 and 2000 depressions. Figure 9a shows the observed vorticity tendencies at 1200 UTC on 10 June 1999 and at 1200 UTC on 5 June 2000, as the depressions were making landfall. Also shown is the calculated vorticity tendency (horizontal advection and divergence only) for both storms. (The magnitude of vertical advection and tilting contributions was very small and was neglected.)

While the observed and calculated vorticity tendencies do not match completely, likely due to subgrid-scale processes such as convection, the analysis shows a clear preference for the NW flank of the depression, which appears to explain the observed motion direction. The horizontal advection of vorticity and the generation of vorticity by convergence were consequences of the wind speed maximum on the NE flanks of these storms (see Fig. 5). This speed maximum is a common feature of monsoon depressions (Rao 1976; Potty et al. 2000). The enhanced convergence on the NW flank is hinted at in Fig. 5, with the deceleration of the flow as it wrapped around the north of each depression. Horizontal advection was most important on the NE flank, where wind speeds were largest. Both storms were traveling to the NW, so wind speeds on the NE flank should be greater than elsewhere since storm motion and storm-relative winds would reinforce each other in that quadrant, and thus maintain the existing direction through vorticity advection and convergence. However, based on tracking the 850-hPa vorticity maxima in the ECMWF data, the 1999 storm averaged a speed of less than 4.5 m s⁻¹, and the 2000 storm's speed was less than 6 m s⁻¹. This is not always enough to resolve the observed discrepancy between wind speeds in different quadrants, particularly at 1200 UTC on 11 June 1999 and at 1200 UTC on 6 June 2000 (Fig. 5).

To assess the potential for blocking to occur, crude estimates of the Froude number (Fr) were calculated for these flows using ECMWF data. Taking a difference of potential temperature Δθ = 30 K between mountaintops and low elevations (ΔH ≅ 5000 m), a reference value of θ₀ = 300 K for the potential temperature at low elevations, and a low-level wind speed U = 10 m s⁻¹, the Froude number is 0.14 [Fr = U/(g ΔθΔH/θ₀)^1/2], which suggests that the low-level flow was blocked by the mountains in both cases. Given the observed instability during the 1999 and 2000 onsets, this result would not apply universally as parcels could be lifted over mountains by convective processes, but it is a good first-order approximation for large-scale flow.

Therefore, as the monsoon depressions leave the Bay of Bengal and move over northern India, the Himalayan range acts as a boundary barrier impeding the flow. The Coriolis acceleration along the mountains vanishes, the pressure gradient normal to the mountains increases, and the along-mountain (easterly) flow is strengthened. Geostrophic adjustment of the monsoon vortex leads to an increase of the absolute angular velocity and centripetal accelerations, thus increasing convergence and cyclonic circulation consistent with the time evolution of these depressions in the infrared imagery (Figs. 7a–b). Because the mountains do not form a uniform obstacle to flow, another form of flow adjustment is to increase shear in response to changes in topography. Nonlinear effects due to the spatial variation of terrain features add complexity to the flow and establish preferential regions of updrafts, which can develop into deep convective cells embedded in the spiral arms of the depression (Gill 1982).

Furthermore, the ECMWF data show low-level cold advection, on the order of several kelvins per day, occurring on the northern flanks of these storms initially. We compare the low-level southeasterly flow to a density current that is colder, and thus denser, than its environment. Blocking of this cold current would contribute to a further increase of the pressure gradient force, accelerating the westward flow along the mountain range, and accentuating the jetlike structure in the northeastern flank of the vortex (Bluestein 1993). These mechanisms are consistent with the observed speed increase on the northeastern flanks of the depressions, especially at times like 1200 UTC on 6 June 2000 (Fig. 5b). Over time friction and latent heat release through precipitation should work to steadily erode the potential energy of the vortex, decelerating the flow, particularly as the depressions got closer to the Himalayas. Thus, the jetlike features should weaken late in the storms' life cycles, as observed.

The agreement between calculated and observed vorticity tendencies did not hold up throughout the duration of the onset storms. Figure 9b shows observed and calculated vorticity tendencies for 0000 UTC on 12 June 1999 and 0000 UTC on 7 June 2000. This is late in the storms' lifetimes, when convective organization and rotation on their northeastern flanks were near maximum. While the observed vorticity tendencies show the storms propagating mostly northward at this point, the calculated tendency is for increasing vorticity on the northeastern flanks. This was almost entirely due to generation by convergence, as the southerly winds on this flank steadily slowed down. As a result of blocking by the Himalayas, the winds in this sector decelerate over time, so that upstream convergence occurred. This region coincides with the area covered by the intense, organized convection in the satellite imagery (Figs. 8a;ndb). It is also a region of positive advection of moisture and warm temperatures, which should work in concert with the vorticity generation to assist in the development of organized convection. Thus, while the ECMWF data are too coarse to resolve any mesoscale vortex generation or convective enhancement on the northeastern flanks, they do provide evidence for why they occurred, even as the depressions overall weakened, filled, and continued to the northwest.

A strong interaction of the seasonal cycle and the 30–60-day global-scale intraseasonal oscillation (ISO; Madden and Julian 1972) with the onset of the Asian summer monsoon was found by many investigators (Murakami et al. 1986; Chen et al. 1988; Kuma 1988; Wang and Xu 1997; Wu et al. 1999). Krishnamurti and Subrahmanyam (1982) showed that the northward-propagating portion of this mode can be associated with the development of cyclonic storms in both the Arabian Sea and Bay of Bengal. Given the timing of the 1999 and 2000 monsoon depressions, it is possible that the ISO played a role in their development, as well as their northward propagation. However, the ECMWF data were not time-filtered to examine this possibility in detail, and no obvious northward motion of vorticity, etc. prior to the development of the depressions was observed in the unfiltered data. In 2001, the depression responsible for the monsoon onset in Nepal (discussed later) developed out of a northward-propagating vorticity maximum, which traveled from near the equator into the Bay of Bengal. However, the subsequent depression traveled initially westward, only turning north several days later. Thus, if this vorticity maximum was associated with the ISO, then it does not appear that the ISO plays a critical role in causing northward propagation of monsoon depressions, at least during 2001.

The influence of storm tracks on onset rainfall is further investigated by an exploratory analysis of central Nepal monsoon onsets in other years (1997, 1998, and 2001). Nepal Department of Hydrology and Meteorology (DHM) daily rainfall data from gauges in the vicinity of the Marsyandi network (1997–1999), Marsyandi network data from June 2001, as well as ECMWF objective analysis data for 1997, 1998, and 2001 were used. The data revealed that early monsoon depressions in the Bay of Bengal also played important roles in the onsets of those years in Nepal, though onset rainfall totals were not as dramatic as 1999 and 2000. While some gauges received comparable rainfall to 1999 and 2000, the breadth of high rainfall totals from those years was not observed in 1997, 1998, or 2001.

Storm tracks for all five depressions are shown in Fig. 10. The 1997 depression formed near where the 1999 depression formed, in the head of the Bay of Bengal. It was about as strong as the 1999 and 2000 depressions but it moved more northward than the 1999 depression, and affected the Himalayas more to the east of the Marsyandi network. The 1998 depression formed south of the 2000 depression and was the weakest of the five. By the time its long trek brought it near the mountains, the resulting southerly flow was significantly weaker than in 1999 or 2000. The 2001 depression formed near the 1997 and 1999 genesis locations, but moved very slowly westward. Based on vorticity advection estimates, its movement probably was impeded by very strong low-level westerlies over India. While the monsoon onset was difficult to detect in the 2001 rainfall data, data from radiosondes launched near the Marsyandi network revealed that the appearance of the monsoon depression was concurrent with a strengthening of southerly to southeasterly winds and an increase in moisture throughout the depth of the troposphere. These radiosondes were launched as part of the Monsoon Himalayan Precipitation Experiment (MOHPREX), which occurred during June 2001. Note that the depressions from 1997, 1998, and 2001 all had some northward component to their motion. Vorticity analyses revealed similar processes to 1999 and 2000 at work.

In all of these cases (1997, 1998, 2001), the tracks of the monsoon depressions, as well as their intensity, precluded the strong southerly flow over the Marsyandi network observed in 1999 and 2000. The tracks also precluded northward advection to the Marsyandi of convection that may have existed on the depressions's eastern flanks. Correspondingly, onset rainfall was not as dramatic as 1999 and 2000, although there was an increase in rainfall during each depression's lifetime, and this increase heralded the arrival of regular rainfall in the subsequent days and weeks (i.e., the monsoon).