mardi 12 janvier 2016

T. S. M. Randriamahefasoa , C. J. C. Reason Interannual variability of rainfall characteristics over southwestern Madagascar

Theoretical and Applied Climatology
pp 1-17

 n'arrive pas à sauvegarder le pdf , à défaut  , page html, compléte (1 seule page ) dispo, copié/collé grossier

Interannual variability of rainfall characteristics over southwestern Madagascar

  • T. S. M. Randriamahefasoa
  • , C. J. C. Reason 


The interannual variability of daily frequency of rainfall [>1 mm/day] and heavy rainfall [>30 mm/day] is studied for the southwestern region of Madagascar, which is relatively arid compared to the rest of the island. Attention is focused on the summer rainy season from December to March at four stations (Morondava, Ranohira, Toliara and Taolagnaro), whose daily rainfall data covering the period 1970–2000 were obtained from the Madagascar Meteorological Service. El Niño Southern Oscillation (ENSO) was found to have a relatively strong correlation with wet day frequency at each station and, particularly, for Toliara in the extreme southwest. In terms of seasonal rainfall totals, most El Niño (La Niña) summers receive below (above) average amounts. An ENSO connection with heavy rainfall events was less clear. However, for heavy rainfall events, the associated atmospheric circulation displays a Southern Annular Mode-like pattern throughout the hemisphere. For ENSO years and the neutral seasons 1979/80, 1981/82 which had large anomalies in wet day frequency, regional atmospheric circulation patterns consisted of strong anomalies in low-level moisture convergence and uplift over and near southwestern Madagascar that made conditions correspondingly more or less favourable for rainfall. Dry (wet) summers in southern Madagascar were also associated with an equatorward (poleward) displacement of the ITCZ in the region.

1 Introduction

Madagascar, the largest island in the Indian Ocean, separated from the African mainland about 170 million years ago and is home to a variety of relatively fragile and unique ecosystems. Oriented south south west to north north east, the island extends roughly 600 km north to south and much less in the zonal direction. Running along most of its meridional extent is a substantial mountain range (Fig. 1) which, due to the moisture-laden easterly trade winds from the South Indian Ocean, leads to orographically enhanced rainfall along the east coast. On the leeward side of the mountains (i.e., in the west and southwest of the country), there is a relative rain shadow effect. The highest peaks exceed 2600 m (Tsaratanana massif in the north, Ankaratra and Andringitra massifs in the centre) with a plateau in the central part of the island ranging in altitude from about 750–1500 m which thus presents a significant obstacle to the low-level trade winds. These easterly trades prevail throughout the year such that the eastern part of the island is rainy in all months with only August and September being somewhat dry (e.g., the station of Taolagnaro in Fig. 2c). Rainfall amounts in eastern Madagascar largely depend on the distance from the coast and the altitude (Pelleray 1954) with several stations receiving on average over 2000 mm per year. By contrast, southwestern Madagascar is much drier, receiving between 300 and 1000 mm annual rainfall.
Fig 1
Map of Madagascar showing the topographic structure and the four rain-gauge stations of Morondava (20S 44E), Ranohira (22S 45E), Toliara (23S 43E) and Taolagnaro (25S 46E). Below the black line is the south western area, the main focus of the study
Fig. 2
Annual cycle of mean monthly rainfall amount (mm) computed for 1971–2000
The northern part of the island experiences a monsoonal climate with the NE monsoon winds over the western tropical Indian Ocean re-curving over the northern Mozambique Channel to become rain-bearing northwesterlies. To its south is the ITCZ which lies over central Madagascar in summer reaching as far south as 17–20° S in February. The mean monthly rainfall amount is 470 mm during summer in northwestern Madagascar (Jury et al. 1995), whereas in the much drier southwest, it ranges from about 50 to 250 mm per month at this time of year (Fig. 2).
Disturbances in the easterly trade winds, including tropical cyclones, are responsible for much of the summer rainfall over the island. On average, the South West Indian Ocean experiences about 9–11 tropical cyclones per season and many of these systems impact on the eastern coast of Madagascar (Mavume et al. 2009). The south west of the island is rarely impacted by tropical cyclones although there are cases like TC Favio in 2007 that tracked close to the south and southwestern coasts (Klinman and Reason 2008) and TC Dera in 2001 (Reason 2007) that tracked southwards through the Mozambique Channel and passed quite close to the southwestern coast of Madagascar. A few tropical cyclones have tracked across the island influencing rainfall (e.g. Eline in 2000, Reason and Keibel 2004). Tropical-extratropical cloudbands, which are large convective systems linking a tropical disturbance over the southern African tropics or the Mozambique Channel to a frontal system passing south of South Africa (e.g. Fauchereau et al. 2009; Manhique et al. 2011; Hart et al. 2013), are also important contributors to seasonal rainfall. As shown by Hart et al. (2013), cloudbands frequently occur over Madagascar in the summer months with a small peak in March before declining—in the 32 seasons analysed, none occur between June and the beginning of October. In addition to cloudbands, the southern part of the island also sometimes experiences cutoff lows (Singleton and Reason 2007; Hart et al. 2013).
Lying to the southwest of southwestern Madagascar on average is the South Indian Convergence Zone (SICZ), a diagonal band of enhanced rainfall extending southeastwards from southeastern Africa out over the midlatitude South West Indian Ocean during summer (e.g. Streten 1973; Cook 2000) that is largely made up of the aggregation of all the cloud band events. Cook (2001) argued that during El Niño events, the SICZ shifts northeastward as a result of an El Niño Southern Oscillation (ENSO)-induced upper level convergence maximum over the equatorial Indian Ocean. Such a shift in the SICZ can also be interpreted as the seasonal aggregation of changes in cloud band intensity and frequency occurring during El Niño (Fauchereau et al. 2009; Hart et al. 2013). Cook (2000, 2001) also suggested that a weakening of the western portion of the South Indian Ocean anticyclone is part of the El Niño-induced impact in the Madagascan region and leads to increased rainfall at least over the central and northern part of the island. This weakening of the anticyclone and development of a cyclonic anomaly southeast of Madagascar during El Niño may be related to ENSO-induced changes in the midlatitude stationary waves in the Southern Hemisphere (Lyon and Mason 2009). However, examination of NCEP re-analyses, satellite rainfall data as well as HadAM3 AGCM simulations forced with observed SSTs to assess the ENSO response over southern Africa/Madagascar (Reason and Jagadheesha 2005) suggests that ENSO impacts over this island are complex and that interpreting these merely as a shift in the SICZ or as a weakening in the South Indian Ocean anticyclone may be inaccurate.
Although there are numerous papers on the rainfall variability of the southern African mainland, rather little work has been done on that of Madagascar itself. Jury et al. (1995) investigated the variability of summer rainfall over central Madagascar, and Nassor et al. (1997) considered extreme rainfall events in the northwest of the island. Almost no work has been done on the relatively dry southwestern region, thus motivating the need for this study. Despite the passage of a few tropical cyclones reaching southern Madagascar which have brought an excess of rain and sometimes caused flooding, water shortages remain a permanent problem here. This region is bordered by the Mozambique Channel to its west and, to its east, the mountainous massif of Ivakoany and the Anosy and Vohimena ranges which reach altitudes of 1000–1500 m (Fig. 1). Given the importance of rain-fed subsistence agriculture to the island, interannual variability in the frequency of wet days during the summer rainy season, heavy rain days and also the associated circulation patterns are studied. These characteristics are chosen because wet or dry spell frequencies are often of more interest and relevance for subsistence farmers of crops like rice and maize than seasonal rainfall totals (Usman and Reason 2004). The importance of wet day frequency for tropical regions has also been emphasised by Moron et al. (2007). In addition, monthly rainfall depth expectations and return periods of rainfall are calculated.

2 Data and methods

Daily rainfall data for 1971–2000 for three stations in the southwest (Morondava 20° 17′ S, 44° 18′ E, Ranohira 22° 33′ S, 45° 24′ E and Toliara 23° 22′ S, 43° 39′ E) were obtained from the Madagascan Meteorological Service. For comparison, data for Taolagnaro (25° 1′ S, 46° 59′ E) were also analysed—this station lies near the southernmost point of the east coast. Consequently, it is far wetter than the other three stations. All of these stations are located at the coast except for Ranohira (Fig. 1) which lies at an altitude of about 825 m and is located some distance on the leeward side of the southern mountain ranges. Data records were complete from 1971 to 2000 for the four stations. Note that these are the only stations in the region that have sufficient daily data for climate variability studies.
Two categories of wet days were considered. Firstly, a day was classified as wet if more than 1 mm of rain was recorded for that day. If more than 30 mm of rainfall was received, then that day was classified as a heavy rainfall day. Such threshold definitions for wet days and heavy rainfall days have been previously used in southern Africa as well as elsewhere in the world (e.g. Weldon and Reason 2014). Examination of the daily data for each station indicated that 30 mm of precipitation recorded in a single day is appropriate to define this day as one of heavy rainfall. Standardised anomalies of the frequency of wet days and in heavy rainfall days were then computed for each summer and correlated with circulation and SST fields. Time series are plotted with the year corresponding to the January of that summer (e.g. the 1979/80 summer is plotted as 1980).
To consider relationships with ENSO, correlations between these standardised anomalies and the Niño 3 and 3.4 indices were performed using NOAA Extended Reconstructed SST (Xue et al. 2003; Smith et al. 2008). Relationships with circulation fields such as 850 hPa geopotential height and omega at 500 hPa were investigated using NCEP re-analysis data (Kalnay et al. 1996). Moisture fluxes were computed at 850 hPa height as the product of wind and specific humidity at this level. Seasons with anomalous numbers of wet or heavy rainfall days were those whose departures exceeded 1 standard deviation from the mean.

3 Results and discussion

To assess rainfall seasonality, Fig. 2 shows the mean monthly rainfall amounts at the four stations. All three of the southwestern stations show a summer peak in rainfall with Toliara, located the furthest south, being the driest. The station near the southernmost tip of the east coast, Taolagnaro, is considerably wetter than the other three stations and has a much less obvious seasonality. Only September is relatively dry at this station, similar to elsewhere on the eastern coast (Pelleray 1954).
Figures 3, 4, 5 and 6 show the fitted rainfall probability plots for each summer month for each station calculated from monthly data for 1960–2010. In some cases, the rainfall amounts on the horizontal axis have been normalised so that the graphical fit is close to linear. The vertical axis shows the probability of exceedance of the corresponding amounts with Table 1 summarising the associated rainfall amounts for exceedances of 25, 50 and 75 %. For example, for Morondava in December of any year, Table 1 indicates that there is a 75 % probability that the rainfall amount for that month will be 52 mm or greater, a 50 % probability that it will be at least 124 mm and a 25 % probability that this station will receive 225 mm or more. By definition, the return period is the reciprocal of the probability expressed as a fraction of the corresponding dependable rainfall; thus, the return period of 225 mm occurring at Morondava in December is 1/0.25 or 4 years. The return periods of larger rainfall amounts can be estimated from the corresponding panel in Figs. 3, 4, 5 and 6. When comparing the monthly rainfall expectations between the four stations, the patterns suggested by the monthly averages plotted in Fig. 2 of Morondava and Ranohira showing strongest seasonality (maximum in January with a strong decrease to the March values), Toliara showing the same behaviour but much reduced amounts and Taolognaro showing the least seasonality but a stronger March than February value are also apparent in Table 1.
Fig. 3
Monthly rainfall probability plot for Morondava (blue crosses are rainfall amounts and the dashed line is the fitted probability distribution)
Fig. 4
Monthly rainfall probability plot for Ranohira (blue crosses are rainfall amounts and the dashed line is the fitted probability distribution)
Fig. 5
Monthly rainfall probability plot for Taolagnaro (blue crosses are rainfall amounts and the dashed line is the fitted probability distribution)
Fig. 6
Monthly rainfall probability plot for Toliara (blue crosses are rainfall amounts and the dashed line is the fitted probability distribution). For March, 14 years received less than 10 mm, which constitute 30 % of the total data. These years were discarded and the fitted (lognormal) distribution is relative to the rest of the data (70 %)

Table 1
Monthly rainfall depth expectations for December, January, February and March for each station
Probability of exceedance (%)
Figure 7 plots the monthly distribution of the frequency of wet days at each station with Toliara and Taolagnaro again showing a much reduced and much increased value, respectively. For Morondava, Ranohira and Toliara, the number of wet days increases significantly in either November or December with peak values in January. During February, there is a small decrease in wet day frequency followed by a sizeable decline in March, the last summer month. From April to October, there are rather few wet days on average. However, for Taolognara, all months show relatively large numbers of wet days except September and there is a small relative peak in January and February and then in April.
Fig. 7
Wet day frequency (number of days per month) computed for 1971–2000
A time series of standardised summer rainfall anomalies for each station is given in Fig. 8 with El Niño (La Niña) summers denoted as red (blue) columns and neutral as unshaded. It is clear that there are several neutral summers at each station that are either substantially wet or dry. However, of the 7 El Niño summers, 5 (Taolagnaro and Toliara) or 6 (Morondava and Ranohira) summers received below average rainfall. For La Niña, 4 out of 6 summers received average or above average rainfall. Thus, Fig. 8 suggests that there may be a link between ENSO and rainfall amount in southwestern Madagascar. Further west, note that a relatively strong ENSO rainfall linkage is found over the southern African mainland (Reason et al. 2000; Reason and Jagadheesha 2005; Ratnam et al. 2014).
Fig. 8
Time series of standardised anomaly in summer rainfall amount for each station. In each case, red corresponds to El Niño and blue to La Niña events
Figure 9 shows the number of wet days and the number of heavy rainfall days at each station through the record, whereas Fig. 10 plots the standardised anomalies in each parameter. It is clear from the former that seasons with high numbers of wet days do not necessarily have high numbers of heavy rain days and vice versa. From Fig. 10, it is evident that most El Niño (La Niña) summers have below (above or near-) average numbers of wet days and that there are also some neutral seasons with large anomalies such as 1979/80 (dry) and 1981/82 (wet). Correlations of the standardised wet day frequency anomaly at each station with various ENSO indicators (SOI and the Niño 3, 3.4 and 4 indices) are shown in Table 2 (only coefficients at above 95 % statistical significance are shown). The strongest correlations are between wet day frequency at Toliara and any of the ENSO indices (r ranges between −0.50 and −0.61 for the SST indices and r = 0.60 for the SOI, all significant at 99 % or better). For the other stations, the correlation coefficients with the various ENSO indices are in the range −0.30 to −0.45.
Fig. 9
Number of days per summer season that are wet (blue part of each column) or receive heavy rainfall (red part of each column)
Fig. 10
Standardised anomalies in wet day frequency. In each case, red corresponds to El Niño and blue to La Niña events

Table 2
Correlation coefficients between the standardised anomaly of the frequency of wet days of each station and various ENSO indices
Nino 1.2
Nino 3
Nino 3.4
Nino 4
Only those correlation coefficients that are significant at the 95 % confidence level are shown
Although there are relatively strong ENSO relationships for wet day frequency, this is not the case for heavy rain day numbers (Fig. 11). Anomalously high frequency in heavy rain days can occur during both El Niño and La Niña events. Figure 11 shows that 1981/82 was a season with an anomalously large number of heavy rainfall days at all stations whereas 1979/80 experienced very few heavy rain days at all stations.
Fig. 11
Standardised anomalies in heavy rain day frequency. In each case, red corresponds to El Niño and blue to La Niña events
Correlations of the standardised anomalies in wet day frequency with global SST (Fig. 12) show an ENSO-like signal for the three southwestern stations, particularly Toliara. In the Indian Ocean, the correlation pattern is also reminiscent of the mature phase of ENSO which typically shows a dipole pattern of one sign in the tropical Indian Ocean and the opposite sign in the southern midlatitudes (Reason et al. 2000). These patterns suggest that local SST does not play an important role in directly modulating the number of wet days; instead, ENSO may change the atmospheric circulation patterns near southwestern Madagascar and thereby influence rainfall. For example, assessment of a number of individual ENSO events in NCEP re-analyses and an atmospheric GCM forced with observed global SST anomalies indicated that changes in the tradewinds near southern Madagascar and associated low-level convergence could be linked to the observed rainfall amounts (Reason and Jagadheesha 2005). For Taolognaro, the ENSO signal is weaker and there is a region of strong negative correlation in the southwest Indian Ocean. This signal is in the opposite sense to what would be expected if local SST were increasing the number of wet days, for example, by increasing the moisture content and instability of the lower atmosphere in the region. Instead, it implies that the changes in wet day frequency at Taolognaro could be caused by regional atmospheric anomalies that also modify the local SST. For heavy rain day frequency (Fig. 13), only Toliara indicates a clear ENSO signal and only Taolagnaro has a correlation signal (−0.2 to −0.3) in the southwest Indian Ocean. For Morondava and Ranohira, the spatial correlations show little in the Indian Ocean or in the tropical Pacific.
Fig. 12
Correlation between wet day frequency anomalies and SST. In each case, coloured areas are statistically significant at 95 %
Fig. 13
Correlation between heavy rain day frequency anomalies and SST. In each case, coloured areas are statistically significant at 95 %
To assess regional atmospheric patterns, the wet day time series was correlated with 850 hPa geopotential height (Fig. 14). Each station shows the general increase (decrease) in geopotential height in the tropics/subtropics across the western Pacific and Indian Oceans associated with El Niño (La Niña) (Reason et al. 2000) which, again, is strongest for Toliara. In addition, there is an area of relative strong negative correlation (implying cyclonic conditions) in the Mozambique Channel/Madagascan region. Such a regional circulation pattern would be favourable for both tropical extratropical cloudbands and tropical storms/cyclones and hence increased wet days and rainfall over southern Madagascar. As implied by Fig. 15 which shows the mean summer moisture flux and its divergence at the 850-hPa level, the cyclonic pattern in the Mozambique Channel/Madagascan region in Fig. 14 effectively reinforces the climatology, thereby indicating a wetter rainy season than average. Correlations between wet day frequency and omega at the 500-hPa level were also performed (Fig. 16). Omega indicates relative uplift or subsidence, and it was found that there was a relatively strong negative correlation over the Mozambique Channel and southern Madagascar. Thus, as expected, there is relative uplift in the midtroposphere for seasons with increased wet day frequency.
Fig. 14
Correlations between wet day frequency anomaly and 850-hPa geopotential height. In each case, coloured areas are statistically significant at 95 %
Fig. 15
Mean DJFM moisture flux and divergence at the 850-hPa level
Fig. 16
Correlations between wet day frequency anomaly and 500-hPa omega (coloured areas are statistically significant at 95 %)
For heavy rain day frequency (Fig. 17), the tropical and ENSO circulation patterns tend to be smaller by comparison to those for wet day frequency, and now, there is a signal reminiscent of the Southern Annular Mode (SAM) (e.g. Thompson and Wallace 2000; Reason and Rouault 2005) for both Morondava and Taolagnaro and, to lesser extent, at Ranohira and Toliara. These patterns suggest that when the SAM is in its positive phase, heavy rainfall over southwestern Madagascar is more likely to occur. In this phase of SAM, the subtropical anticyclones shift southward with easterly wind anomalies in the subtropics. The high-pressure anomalies in the midlatitude region near South Africa associated with this SAM-like pattern are located such to imply increased easterly wind flow of moist unstable air from the South West Indian Ocean towards southern Madagascar. This feature, together with weak negative correlations in the southern Mozambique Channel and southern Madagascar which imply relative cyclonicity there, are favourable for heavy rainfall.
Fig. 17
Correlations between heavy rain day frequency anomaly and 850-hPa geopotential height. In each case, coloured areas are statistically significant at 95 %
The correlation patterns for omega at 500 hPa (not shown) are negative over Madagascar, the Mozambique Channel and neighbouring South West Indian Ocean, consistent with heavy rainfall over the southern part of the island.

4 Case studies—non-ENSO anomalous years: 1979/80 and 1981/2

Figures 8 and 10 show that there are some neutral seasons that were substantially wet or dry and with large anomalies in wet day frequency implying that ENSO is not the only factor of importance. For improvements in the understanding of regional climate variability and the potential of seasonal forecasting, it is therefore useful to investigate the circulation patterns associated with neutral seasons with large anomalies in rainfall characteristics. As case studies, this section considers the neutral summers of 1979/80 and 1981/82 which stand out as ones with anomalously low and high numbers of wet days, respectively, at the three southwestern stations as well as in heavy rain days (Fig. 11). There is also anomalously low heavy rain day frequency at Taolagnaro in 1979/80 and the reverse in 1981/82 (Fig. 11c). Figure 18 shows the 850-hPa moisture flux anomaly over the region for 1979/80 indicating a large cyclonic anomaly and moisture convergence east of Madagascar with relative divergence over the southwest of the island. Globally, the 850-hPa geopotential height anomaly displayed a weak El Niño-like pattern (not shown) with weak positive height anomalies over southern Africa and southwestern Madagascar as well as the maritime continent region (not shown). The ITCZ was weaker over the tropical South West Indian Ocean and southeastern Africa during this season. Furthermore, all of western Madagascar, and particularly the south west, shows strong relative subsidence which extends across the Mozambique Channel and most of southern Africa (Fig. 19). This pattern is very unfavourable for cloud band occurrence across the region or tropical storm development in the Mozambique Channel consistent with the much reduced number of wet days and drier conditions in this season. Furthermore, there were large cool SST anomalies over the southwest Indian Ocean/greater Agulhas Current system south of Madagascar (not shown), a pattern unfavourable for summer rainfall in southeastern Africa/southern Madagascar (Reason and Mulenga 1999; Reason 2002).
Fig. 18
Anomalies in moisture flux (vectors) and divergence (contours) for DJFM 1979/80
Fig 19
Anomalies in omega at the 500-hPa level for DJFM 1979/80. Positive (negative) values indicate relative sinking (uplift)
During the 1981/82 summer, which had anomalously high wet day and heavy rain day frequencies, the large-scale 850-hPa geopotential height (not shown) displayed weak negative anomalies in the maritime continent region and a wave number 3 pattern in the Southern Hemisphere with a strong anticyclonic anomaly to the southeast of Madagascar which would enhance the advection of moist marine air towards the region. Further north, a large cyclonic anomaly established itself across the southern Mozambique Channel from eastern Zimbabwe to eastern Madagascar (Fig. 20). The strong westerly anomalies on the eastern side of the channel led to substantial low-level moisture convergence over southwestern Madagascar that extended to a second cyclonic anomaly to the south east of the island. Coupled with this pattern was a region of strong relative uplift in the midtroposphere that extended from south east of Madagascar across the southwest of the island towards the central part of the channel (Fig. 21). These circulation anomalies were favourable for the development of cloud bands and tropical storms across the southern part of the channel and southwestern Madagascar leading to increased rainfall and wet day frequency during this summer. SST anomalies during 1981/82 summer showed a positive subtropical South Indian Ocean dipole pattern, favourable for increased rainfall over southeastern Africa and southwestern Madagascar (Behera and Yamagata 2001; Reason 2002).
Fig 20
Anomalies in moisture flux (vectors) and divergence (contours) for DJFM 1981/82
Fig 21
Anomalies in omega at the 500-hPa level for DJFM 1981/82. Positive (negative) values indicate relative sinking (uplift)
It is of interest to note that the dry 1979/80 and wet 1981/82 summers over southwestern Madagascar had, respectively, more and less cloudbands over the southern African mainland than average (Hart et al. 2013) and were also, respectively, wetter and drier over large parts of Namibia, northern South Africa, Botswana and Zimbabwe. To our knowledge, these types of opposing rainfall signals between subtropical southern Africa and southwestern Madagascar have not been noted before. Furthermore, there were less than average numbers of tropical storms in the Mozambique Channel in 1979/80 but an average number in 1981/82 (Mavume et al. 2009) consistent with the former season being drier over southwestern Madagascar with less rain days and the opposite for 1981/82.

5 Conclusion

The climate variability of Madagascar has not been much studied, especially the southwest. This region of the island is much drier than the eastern and northern parts as it lies in a rain shadow due to its location on the leeward side of the mountains that run along the island and which present an obstacle to the prevailing easterly trade wind flow.
Given that much of the population of the island depends on rain-fed agriculture, focus was placed on the interannual variability of the number of wet days and heavy rain days in the summer (December to March) rainy season. In subsistence agriculture, a rainy season that experiences regular rainfall (high wet day frequency) is more beneficial than one with a large seasonal rainfall total but infrequent and irregular rainfall (high heavy rain day frequency but relatively low wet day frequency) (Usman and Reason 2004). Four rainfall stations were considered, three in the southwest of the island and the other at the southern extremity of the east coast for comparative purposes. These were the only stations in southwestern Madagascar that had sufficient daily data for climate variability analysis. A wet day was defined as one with at least 1 mm of recorded rainfall whereas a heavy rainfall day was one experiencing at least 30 mm of rainfall. The rainy season begins in December and ends in March for the three southwestern stations (Morondawa, Ranohira and Toliara), whereas almost the entire year, except perhaps September, is wet at Taolagnaro as it lies on the windward side of the mountains and is impacted by moist easterly trade winds throughout the year. Another topographic effect is seen at Ranohira which, although lying polewards of Morondawa, is in fact wetter since the mountains upstream of this station are lower and narrower in zonal extent than for Morondawa.
It was found that most summers with an anomalously high (low) number of wet days occurred during La Niña (El Niño) events whereas there was a less obvious link between heavy rainfall days and ENSO. In general, summer rainfall amounts tend to be more (less) during La Niña (El Niño) years. The strongest correlations with ENSO at wet day frequency were found for Toliara (|r|∼0.5–0.6) although the correlations at the other stations are still significant at the 95 % level. However, there were also some notable neutral summers such as 1979/80 and 1981/82 that experienced well below (above) average numbers of wet days and reduced (increased) rainfall amount. Although there were no obvious ENSO linkages with heavy rain day frequency, a Southern Annular Mode (SAM)-type pattern was clearly present for two of the stations with the other two stations showing these patterns to a weaker extent.
In all cases, the circulation patterns associated with summers with anomalous wet day frequency consisted of strong anomalies in regional low-level moisture convergence and uplift that made conditions correspondingly more or less favourable for rainfall. Dry (wet) summers in southwestern Madagascar were also associated with an equatorward (poleward) displacement of the ITCZ in the region.


The first author is grateful to Carnegie WIO-RISE for the funding of her MSc research from which this manuscript is written.

Aucun commentaire: