Cape Verde Islands are located off the West African coast between 14˚N - 18˚N and 22˚W - 28˚W in North Atlantic Ocean. The islands lie 570 km off the Senegal coast and constitute an archipelago of volcanic origin, composed by ten islands-nine inhabited islands and one uninhabited-which together have a 4.033 km² land cover surface. According to [1] , it is inserted on the Sahel climate zone-arid and semi-arid-crossing Africa from Atlan- tic to Red Sea, and it extends from Saudi, Syria and Mesopotamia, to temperate deserts zone of Eurasia and sub- desert regions. These areas are strongly affected by monsoon climates.

Air temperature is around 20˚C in most of country in coldest period from January to April and reaches values higher than 27˚C in hottest period, from August to October. Annual rainfall varies from less than 100 - 150 mm on arid flat coastal to around 800 - 900 mm in highest mountainous areas above 1000 m elevation. Cape Verde climate is governed by a strong diurnal cycle and the annual cycle by West African monsoon circulation with two different wind regimes [2] .

With regard to systems that affect the archipelago’s climate, it is noteworthy the variable wind regime through- out the year. Northeast trade winds predominate, blowing strongly between December and May. Past studies suggest that Intertropical Convergence Zone (ITCZ) affects directly Cape Verde’s climate. ITCZ presence over Cape Verde’s latitudes is, however, negatively affected by pressure fluctuations of Azores anticyclone and other high altitude air mass fluxes in northern Central Atlantic [3] . An extremely variable rainfall regime results from oscillations of this regional high and low pressure zones. The archipelago may be affected by squall lines or tropical disturbances that originate over the African Continent. These phenomena are well describ- ed in [4] .

West Africa rainfall, on the other hand, has been subject of many studies and research due to the location of various countries in the Sahel region and the extreme variability of precipitation in those areas. Sahel, extending across Africa from Atlantic Ocean to Indian Ocean, is the largest semiarid region of world. Studies vary from seasonal systems of active precipitation to seasonality (i.e., duration of the dry and wet season) and possible trends of these variations.

A classical and dominant feature of atmospheric dynamics in the tropics is the ITCZ, characterized by a belt of low-level trade winds convergence and upper-level divergence with strong updrafts, deep convection, fre- quent cloudiness, and heavy rainfall [5] . Past and modern studies have suggested an influence of the ITCZ in the rainfall over West Africa (e.g., [6] - [8] ).

A study by [9] computed radiosonde data averages for 11 year period (1960-70) of eight stations located along latitude of 5˚N representing the monsoon belt, and others around 13˚N - 14˚N, a typical zone of semiarid climatic regime (i.e., southern and northern zones, respectively). Authors mentioned that northern climate is af- fected by the surface ITCZ’s excursion and the intrusion of mid-latitude westerlies.

Annual rainfall long-term series for 69 stations throughout the entire African continent including a Cape Verde station—Sao Vicente Island (Mindelo at 16˚53'N, 25˚00'W)—were examined [10] . Results show a gener- ally oscillatory trend at beginning followed by a negative trend. He found that, overall, no significant trends or periodicities were evident, although the oscillatory characteristic is present in most data analyzed.

It was pointed out that, during boreal summer when ITCZ moves northward with increasing solar declination, a heat low is generated over Sahara Desert. Moist southwesterly winds at surface and African easterly jet in middle troposphere are dominant over the continent [8] .

The rainfall variability in the Western Sahel was studied using a long-term dataset in the period between 1950 and 2004 for 14 stations of 6 countries in West Africa Sahel area (Niger, Mali, Mauritania, Burkina Faso, Benin, and Senegal-located in close proximity of Cape Verde) [11] . The results present solid evidence associating the rain regime to the period the ITCZ reaches its maximum position in its movement to the north, and more intense activity of a classical SW monsoon. The study also indicates that oscillatory movement of the ITCZ is a signa- ture of the same phenomenon for the dry regime when the ITCZ migrates to the south, and its influence is expe- rienced over the transition regime.

The Cape Verde population has increased 40% over the past 20 years, reaching about half million inhabitants, and definitely it will continue to increase in the near future. Water resources, both for human and animal con- sumption, as well as crop cultivation, are scarce and already a limiting factor for society welfare at present. As opposed to West Africa, the Cape Verde rainfall regime, and its relationship with large scale atmospheric phe- nomena, has never been studied in details. Due to its geographical location and proximity to the Sahel, it always has been supposed to behave like West Africa’s. The Santiago Island is selected for this case study since it has available a reliable 29-year period rainfall data set of four gauges of the Cape Verde National Institute of Geo- physics and Meteorology (INMG, in Portuguese) network and it is more vulnerable because it holds more than 50% of the country’s population. These observed data sets in conjunction with the NCEP/NCAR reanalysis data set are used to characterize its rainfall regime and to identify apparent climate cycles and possible large scale atmospheric mechanisms that are responsible for its rainfall production, in particular the hypothesis that the ITCZ is the main rainfall producing mechanism. It is expected that the results presented here will contribute to better rainfall forecasting, both in the short and long ranges, that, in turn, will help to develop and implement policies and strategies for water resources usage in the Archipelago.

The analysis is presented as follows. Section 2 describes the data and methods used in this investigation. A descriptive characterization of rainfall patterns of Santiago over the study period is presented in Section 3, ex- amining long-term seasonality and possible rain-producing systems. Summary and conclusion follow in Section 4.

Santiago (14˚54' - 15˚20'N; 23˚25' - 23˚46'W) is the largest island in the archipelago and its most significant agricultural center with an area of 991 km². According to their geopolitical characteristics, the Island is divided into 6 key municipalities, as shown in Figure 1. The Island is characterized by scarce hydrologic resources due to the low and irregular rainfall totals (average 321 mm/year in the highest mountains and about 169 mm/year at low levels) and also that this fact is responsible for some quasi-periodic and catastrophic dryness. The small amount of superficial water combined with irregular precipitation explains the strong dependence of the popula- tion on subterranean water [12] . Studies on water balance indicate that runoff takes 18% of precipitation and it is directly discharged to the sea, and approximately 13% infiltrates the soil. The remaining water is lost by evapo- transpiration, which attains a mean annual potential value of 1500 mm [13] . Despite the low rainfall values, very intensive rain events, together with the topography effect, lead to flood at most of the existent creeks [14] .

Praia City is the capital of Cape Verde and concentrates most of the country’s population. Santiago’s climate is conditioned by its geomorphology [15] . As a result of the altitude, it is noted that, moving within the Island, the arid coastal zone climate type becomes semi-arid, and finally sub-humid, as shown in Table 1. These cha- racteristics strongly define the main agricultural products (e.g., corn, sugar cane, banana, coffee and mangoes).

Rainfall data sets from four stations of the INMG were used for this analysis. The length of the records is the same for all rainfall series examined. Santiago Island long-term rainfall data permitted a regional analysis to examine the variation in the rainfall patterns for the 29-year period 1981-2009. Data records were from 3 me- teorological stations and 1 rain gauge station set up according to the World Meteorological Organization stan- dards. Figure 1(c) displays the geographical distribution of weather stations used and Table 2 summarizes the

¹Meteorological station; ²Rain gauge station.

stations’ information.

It has been common practice to assume that at least ten years of daily observations are necessary to produce the relevant base statistical parameters for most meteorological elements, and at least thirty years for precipita- tion [16] . Simple descriptive statistics for measured monthly, seasonal, and annual rainfall were calculated for the period from 1981 to 2009 for the gauges with continuous records. To summarize the rainfall changes over the study period, a rainfall index was also used by standardizing the time series of four stations with the 29 yr of data according to [17] :

where x_i,j is the standardized monthly rainfall deviations; p_i,j is the annual rainfall for the Island; μ_i is the mean value of 1981-2009 period; and the σ_i the standard deviation 1981-2009 period. This index allows to highlight driest and wettest years over all the period. Time versus longitude plots of zonal (U) and meridional (V) wind components anomalies (Operational Data Plot—2.5˚ × 2.5˚ resolution) for the 850 hPa level over the region are made using reanalysis data from ESRL/PSD/NOAA website. The variables are averaged over 5˚ latitudinal strip, from 12.5˚N - 17.5˚N, and the longitudinal range of 0˚W to 60˚W is in the horizontal axis. These anomalies were used to observe the wind pattern dominance during dry and wet years (i.e., humid winds from Atlantic Ocean and dry air from Sahara desert region). An accurate description of the NCEP/NCAR Reanalysis project can be found in [18] . A large subset of these data is available from ESRL in its original format of 6 hour inter- vals and daily average analyses. The project uses a state-of-the-art analysis/forecast system to perform data as- similation using past data from 1948 to the present. These data are selected because of it is the best data set available at present moment, for its free availability and easy access.

Data of daily outgoing longwave radiation anomalies (OLR) are often used asproxy of areas of tropical con- vection and to estimate the earth’s radiation budget. In this study, daily OLR time versus longitude section plots are made to track possible weather systems or traveling disturbances over Santiago Island. OLR data are aver- aged for latitudinal belt 12.5˚N - 17.5˚N and extend from 0˚W to 60˚W of longitude. For more details about OLR anomalies description read [19] , for example. Time sections of sea level pressure (SLP) are also elaborated over the same geographic domain as a support for the OLR anomaly analysis. The OLR and SLP data are taken from ESRL/PSD/NOAA. Time section plots are made with the applets existing in the same website. Lastly, 2 years, the driest year (1994) and the wettest one (1999) in the record are selected for analyzing the atmospheric phenomena that generate wet years and dry years.

This paper presents results of long-term analysis of rainfall data collected at the Santiago Island weather stations network of Cape Verde National Institute of Meteorology and Geophysics. The datasets of four stations are analyzed for the period 1981-2009, 29-year records. Rainfall distribution, seasonality and long-term variability are analyzed using monthly and annual averages relative to the study period.

Rainfall over Cape Verde is concentrated in 3 to 4 months (July-October), especially in August and September which, together, are responsible for more than 75% of the total precipitation. September exhibits the highest accumulated precipitation. The dry season is from November to June (8 months) and it is characterized by sporadic rainfall events, mostly from November to January, discussed as transition months when remnants of cold fronts reach the archipelago. During the transition months, precipitation values are, generally, less than 15 mm/month. These results clearly show that Santiago has well defined rainfall regime, which is similar to semi- arid climate regions. There is a strong evidence of direct influence of orography over the total precipitation, a result in agreement with previous studies. It is worth to point out that there are single months of extreme

drought within the 29-year period studied.

The OLR anomalies and the SLP anomalies analysis, used as proxies for weather systems propagation, show that most of rain-producing systems that affect Cape Verde are associated with transient convective disturbances. Evidently, traveling disturbances are more frequent and intense during wet years than dry years. Time versus longitude section plots of U and V wind components anomalies for 1994, the driest year, indicate that the V anomalies remained negative and weak for most of the rainy season and occasional positive surges of southwest- terly winds were associated with OLR and SLP negative anomalies. In contrast, during 1999 wet year, strong and intense winds, blowing from the southwestern quadrant over Atlantic bring moisture, enhances the conver- gence flux and, thus, rainfall over the Archipelago. So, these results indicate that rainfall is produced by south- westerly transient disturbances associated with the frontal systems that cross the Atlantic and not with ITCZ or squall lines coming from North Africa.

Standardized rainfall index proved to be an excellent method to identifying the extreme years within the pe- riod. The results show that there is strong inter-annual rainfall variability in Cape Verde. However, apparently it is not related with El Niño-Southern Oscillation (ENSO) phenomenon and its cause must be further investigated. This result is a surprise, as West Africa rainfall presents some relation with ENSO (e.g., [21] [22] ). The rainfall series was broken into two periods. The first period 1981-1998 coincides with the PDO warm phase. This period has only two years, 1986 and 1987, with rainfall totals above one standard deviation. The other 16 years (89%), including the driest year, have rainfall either below or near the average. The second period, 1999-2009, is cha- racterized by 7 rainy years out of 11 years, in particular the rainiest year in the whole period. 1999. Only one year was moderately dry (2002), but its departure module is below one standard deviation. In other words, there are no drought conditions during the present PDO cold phase. These results suggest an inter-decadal rainfall va- riability that may be related to PDO. The North Atlantic Oscillation (NAO) may also be related to the Cape Verde rainfall, but its relationship was not investigated and it may be object of future study. Considering that PDO

phases last 25 - 30 years and the current phase may go further another 10 to 15 years, it is possible that the se- vere drought years may decrease in frequency in this near future.

The understanding the genesis of these systems will certainly improve weather and climate forecasting, both short and seasonal term, contributing to developing and implementing policies and strategies for water resources usage in Cape Verde.

The authors acknowledge the contribution of the National Institute of Meteorology of Cape Verde (INMG) for providing the precipitation data. G. E. V-L thanks the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior—CAPES for the financial support.