• AGRHYMET, 1996: Méthodologie de suivi des zones à risque. AGRHYMET FLASH, Bulletin de Suivi de la Campagne Agricole au Sahel 0/96, Vol. 2., 2 pp. [Available from Centre Regional AGRHYMET, B.P. 11011, Niamey, Niger.].

  • Ati, F. O., , C. J. Stitger, , and E. O. Oladipo, 2002: A comparison of methods to determine the onset of the growing season in northern Nigeria. Int. J. Climatol., 22 , 731742.

    • Search Google Scholar
    • Export Citation
  • Benoit, P., 1977: The start of the growing season in northern Nigeria. Agric. Meteor., 18 , 9199.

  • Camberlin, P., , S. Janicot, , and I. Poccard, 2001: Seasonality and atmospheric dynamics of the teleconnection between African rainfall and tropical sea-surface temperature: Atlantic vs. ENSO. Int. J. Climatol., 21 , 9731005.

    • Search Google Scholar
    • Export Citation
  • Chmielewski, F. M., , and T. Rötzer, 2002: Annual and spatial variability of the beginning of growing season in Europe in relation to air temperature changes. Climate Res., 19 , 257264.

    • Search Google Scholar
    • Export Citation
  • Cook, K. H., 2001: A Southern Hemisphere wave response to ENSO with implications for southern Africa precipitation. J. Atmos. Sci., 58 , 21462162.

    • Search Google Scholar
    • Export Citation
  • Dent, M. C., , S. D. Lynch, , and R. E. Schulze, 1989: Mapping mean annual and other rainfall statistics over southern Africa. Water Research Commission Tech. Rep. 109/1/89, Pretoria, South Africa, 230 pp.

  • Eltahir, E., , and C. Gong, 1996: Dynamics of wet and dry years in West Africa. J. Climate, 9 , 10301040.

  • Fischer, G., , H. van Velthuizen, , F. Nachtergaele, , and S. Medow, cited. 2004: Global agro-ecological zones—2000. Food and Agriculture Organisation of the United Nations and International Institute for Applied Systems Analysis. [Available online at http://www.fao.org/WAICENT/FAOINFO/AGRICULT/AGL/agll/gaez/index.htm.].

  • Harrison, M. S. J., 1984: A generalized classification of South African summer rain-bearing synoptic systems. J. Climatol., 4 , 547560.

    • Search Google Scholar
    • Export Citation
  • Hewitson, B. C., , and R. G. Crane, 2002: Self Organizing Maps: Applications to synoptic climatology. Climate Res., 22 , 1326.

  • Hulme, M., 1992: Rainfall changes in Africa: 1931–1960 and 1961–1990. Int. J. Climatol., 12 , 685699.

  • Hulme, M., 1996: Recent climatic change in the world’s drylands. Geophys. Res. Lett., 23 , 6164.

  • International Research Institute for Climate Prediction, 2001: Proceedings, Communication of Climate Forecast Information Workshop. International Research Institute for Climate Prediction, 51 pp.

  • Janicot, S., , A. Harzallah, , B. Fontaine, , and V. Moron, 1998: West African monsoon dynamics and eastern equatorial Atlantic and Pacific SST anomalies (1970–88). J. Climate, 11 , 18741882.

    • Search Google Scholar
    • Export Citation
  • Jury, M. R., 1996: Regional teleconnection patterns associated with summer rainfall over South Africa, Namibia and Zimbabwe. Int. J. Climatol., 16 , 135153.

    • Search Google Scholar
    • Export Citation
  • Kiladis, G. N., , and K. C. Mo, 1999: Interannual and intraseasonal variability in the Southern Hemisphere. Meteorology of the Southern Hemisphere, Meteor. Monogr., No. 49, Amer. Meteor. Soc., 307–336.

  • Kowal, J. M., , and D. T. Knabe, 1972: An Agroclimatological Atlas of the Northern States of Nigeria. Ahmadu Bello University Press, 15 pp.

    • Search Google Scholar
    • Export Citation
  • Lamb, P. J., , and R. A. Peppler, 1992: Further case studies of tropical Atlantic surface atmospheric and oceanic patterns associated with sub-Saharan drought. J. Climate, 5 , 476488.

    • Search Google Scholar
    • Export Citation
  • Landman, W. A., , and W. J. Tennant, 2000: Statistical downscaling of monthly forecasts. Int. J. Climatol., 20 , 15211532.

  • Lizaso, J. I., , and J. T. Ritchie, 1997: Maize shoot and root response to root zone saturation during vegetative growth. Agron. J., 89 , 125134.

    • Search Google Scholar
    • Export Citation
  • Makarau, A., , and M. Jury, 1997: Seasonal cycle of convective spells over southern Africa during austral summer. Int. J. Climatol., 17 , 13171332.

    • Search Google Scholar
    • Export Citation
  • Mason, S. J., 1996: Rainfall trends over the Lowveld of South Africa. Climatic Change, 32 , 3554.

  • Matarira, C. H., 1990a: Frequency and tracks of anticyclones and their effect on rainfall patterns over Zimbabwe. Theor. Appl. Climatol., 42 , 5366.

    • Search Google Scholar
    • Export Citation
  • Matarira, C. H., 1990b: Drought over Zimbabwe in a regional and global context. Int. J. Climatol., 10 , 609625.

  • Matarira, C. H., , and M. R. Jury, 1992: Contrasting meteorological structure of intra-seasonal wet and dry spells in Zimbabwe. Int. J. Climatol., 12 , 165176.

    • Search Google Scholar
    • Export Citation
  • Mo, K. C., 2000: Relationships between low-frequency variability in the Southern Hemisphere and sea surface temperature anomalies. J. Climate, 13 , 35993610.

    • Search Google Scholar
    • Export Citation
  • Nicholson, S. E., 1994: Recent rainfall fluctuations in Africa and their relationships to past conditions over the continent. Holocene, 4 , 121131.

    • Search Google Scholar
    • Export Citation
  • Nicholson, S. E., 1995: Variability of African rainfall on interannual and decadal time scales. Natural Climate Variability on Decade-to-Century Time Scales, Climate Research Committee SE, National Academy Press, 32–43.

    • Search Google Scholar
    • Export Citation
  • Nicholson, S. E., , B. Some, , and B. Kone, 2000: An analysis of recent rainfall conditions in West Africa, including the rainy seasons of the 1997 El Niño and the 1998 La Niña years. J. Climate, 13 , 26282640.

    • Search Google Scholar
    • Export Citation
  • Olaniran, O. J., 1983: The onset of the rains and the start of the growing season in Nigeria. Niger. Geogr. J., 26 , 8188.

  • Omotosho, J. B., 1990: Onset of thunderstorms and precipitation over northern Nigeria. Int. J. Climatol., 10 , 849860.

  • Omotosho, J. B., 1992: Long-range prediction of the onset and end of the rainy season in the West African Sahel. Int. J. Climatol., 12 , 369382.

    • Search Google Scholar
    • Export Citation
  • Omotosho, J. B., , A. A. Balogun, , and K. Ogunjobi, 2000: Predicting monthly and seasonal rainfall, onset and cessation of the rainy season in West Africa using only surface data. Int. J. Climatol., 20 , 865880.

    • Search Google Scholar
    • Export Citation
  • Press, W. H., , B. P. Flannery, , S. A. Teukolsky, , and W. T. Vetterling, 1993: Numerical Recipes in C: The Art of Scientific Computing. Cambridge University Press, 994 pp.

    • Search Google Scholar
    • Export Citation
  • Sarria-Dodd, D. E., , and I. T. Jolliffe, 2001: Early detection of the wet season in semiarid tropical climates of western Africa. Int. J. Climatol., 21 , 12511262.

    • Search Google Scholar
    • Export Citation
  • Sivakumar, M. V. K., 1988: Predicting rainy season potential from the onset of rains in the southern Sahelian and Sudanic climatic zones of West Africa. Agric. For. Meteor., 42 , 295305.

    • Search Google Scholar
    • Export Citation
  • Smale, M., , and T. S. Jayne, 2003: Maize in Eastern and Southern Africa: “Seeds” of success in retrospect. EPTD discussion paper 97, International Food Policy Research Institute, 79 pp.

  • Stern, R. D., , M. D. Dennett, , and D. J. Garbutt, 1981: The start of the rains in West Africa. J. Climatol., 1 , 5968.

  • Tadross, M., , B. C. Hewitson, , and M. Usman, 2003: Calculating the onset of the maize growing season over southern Africa using GTS and CMAP data. CLIVAR Exchanges, No. 27, International CLIVAR Project Office, Southampton, United Kingdom, 48–50.

  • Tennant, W. J., 2002: Event characteristics of intra-seasonal climate circulations. Ph.D. thesis, University of Cape Town, 185 pp.

  • Tennant, W. J., , and B. C. Hewitson, 2002: Intra-seasonal rainfall characteristics and their importance to the seasonal prediction problem. Int. J. Climatol., 22 , 10331048.

    • Search Google Scholar
    • Export Citation
  • Todd, M. C., , and R. Washington, 1999: Circulation anomalies associated with tropical-temperate troughs over southern Africa and the southwest Indian Ocean. Climate Dyn., 15 , 937951.

    • Search Google Scholar
    • Export Citation
  • Tyson, P. D., , and R. A. Preston-Whyte, 2000: The Weather and Climate of Southern Africa. 2d ed. Oxford University Press, 396 pp.

  • Usman, M. T., 1999: A convective instability index for summer monsoon onset determination over the Sahel. Proc. Workshop on the West African Monsoon Variability and Predictability (WAMAP), Dakar, Senegal, WMO.

  • Usman, M. T., , and C. J. C. Reason, 2004: Dry spell frequencies and their variability over southern Africa. Climate Res., 26 , 199211.

  • Usman, M. T., , E. Archer, , P. Johnston, , and M. Tadross, 2004: A conceptual framework for enhancing the utility of rainfall hazard forecasts for agriculture in marginal environments. Nat. Hazards, 34 , 111129.

    • Search Google Scholar
    • Export Citation
  • Venegas, S. A., 2003: The Antarctic Circumpolar Wave: A combination of two signals? J. Climate, 16 , 25092525.

  • Vogel, C. H., 1994: (Mis)management of droughts in South Africa: Past, present, future. South Afr. J. Sci., 86 , 382386.

  • Walter, M. W., 1967: Length of the rainy season in Nigeria. Niger. Geogr. J., 10 , 123128.

  • Washington, R., , and M. Todd, 1999: Tropical-temperate links in southern African and southwest Indian Ocean satellite-derived daily rainfall. Int. J. Climatol., 19 , 16011616.

    • Search Google Scholar
    • Export Citation
  • Xie, P., , and P. A. Arkin, 1997: Global precipitation: A 17-year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs. Bull. Amer. Meteor. Soc., 78 , 25392558.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Mean number of stations contributing to the rainfall in each gridded bin: (a) CCWR (1950–97) and (b) GTS (1979–2001); with an average of 3 to 5 stations per region.

  • View in gallery

    Mean onset of the maize growing season (number of pentads after 3 Aug) for the 1979–97 period: (a) CMAP and (b) CCWR.

  • View in gallery

    Std dev of onset (number of pentads) for the period 1979–97: (a) CMAP and (b) CCWR.

  • View in gallery

    Trend in onset (pentads yr−1) for the period 1979–97: (a) CMAP and (b) CCWR.

  • View in gallery

    CCWR trend in onset (pentads yr−1) for the 1950–97 period. Contours indicate height of local topography (m).

  • View in gallery

    Time series of onset for the Limpopo, FS, Kwazulu–Natal (KZN), and northeast (28.5°–22.5°S, 25°–33°E) regions of South Africa.

  • View in gallery

    Archetype pentad rainfall patterns associated with each node of the 6 × 4 SOM. Scale is in mm day−1.

  • View in gallery

    Frequency anomaly of late-minus-early distribution of rainfall patterns (SOM nodes) in the six pentads during onset for four regions: (a) FS, (b) SZ, (c) CZ, and (d) WZ. Probabilities (p) are that the late and early frequency distributions of rainfall patterns are similar according to a χ2 test.

  • View in gallery

    Same is in Fig. 8, but during Aug for the four regions.

  • View in gallery

    Daily 500-hPa eddy geopotential height anomalies (m) associated with each node of the 4 × 3 SOM.

  • View in gallery

    Frequency anomaly of late-minus-early distribution of daily 500-hPa eddy geopotential patterns (SOM nodes) in the 30 days during onset for four regions: (a) FS, (b) SZ, (c) CZ, and (d) WZ. Late and early frequency distributions are different at the 99% confidence level.

  • View in gallery

    Same as in Fig. 11, but during Aug.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 121 121 23
PDF Downloads 83 83 22

The Interannual Variability of the Onset of the Maize Growing Season over South Africa and Zimbabwe

View More View Less
  • 1 Climate Systems Analysis Group, Department of Environmental and Geographical Sciences, University of Cape Town, Rondebosch, South Africa
  • | 2 Department of Geography, Federal University of Technology, Minna, Nigeria
© Get Permissions
Full access

Abstract

Subsistence farmers within southern Africa have identified the onset of the maize growing season as an important seasonal characteristic, advance knowledge of which would aid preparations for the planting of rain-fed maize. Onset over South Africa and Zimbabwe is calculated using rainfall data from the Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP) and the Computing Center for Water Research (CCWR). The two datasets present similar estimates of the mean, standard deviation, and trend of onset for the common period (1979–97) over South Africa. During this period, onset has been tending to occur later in the season, in particular over the coastal regions and the Limpopo valley. However, the CCWR data (1950–97) indicate that this is part of long-term (decadal) variability.

Characteristic rainfall patterns associated with late and early onset are estimated using a self-organizing map (SOM). Late onset is associated with heavier rainfall over the subcontinent. When onset is early over Zimbabwe, there is an increased frequency of more intense rainfall over northeast Madagascar during the preceding August. Accompanying these intense events is an increased frequency of positive 500-hPa geopotential height anomalies to the southeast of the continent. Similar positive height anomalies are also frequently present during early onset. The study indicates that onset variability is partly forced by synoptic conditions, and the successful use of general circulation models to estimate onset will depend on their simulation of the zonally asymmetric component of the westerly circulation.

Corresponding author address: Dr. M. Tadross, Climate Systems Analysis Group, Department of Environmental and Geographical Science, University of Cape Town, Rondebosch 7701, South Africa. Email: mtadross@egs.uct.ac.za

Abstract

Subsistence farmers within southern Africa have identified the onset of the maize growing season as an important seasonal characteristic, advance knowledge of which would aid preparations for the planting of rain-fed maize. Onset over South Africa and Zimbabwe is calculated using rainfall data from the Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP) and the Computing Center for Water Research (CCWR). The two datasets present similar estimates of the mean, standard deviation, and trend of onset for the common period (1979–97) over South Africa. During this period, onset has been tending to occur later in the season, in particular over the coastal regions and the Limpopo valley. However, the CCWR data (1950–97) indicate that this is part of long-term (decadal) variability.

Characteristic rainfall patterns associated with late and early onset are estimated using a self-organizing map (SOM). Late onset is associated with heavier rainfall over the subcontinent. When onset is early over Zimbabwe, there is an increased frequency of more intense rainfall over northeast Madagascar during the preceding August. Accompanying these intense events is an increased frequency of positive 500-hPa geopotential height anomalies to the southeast of the continent. Similar positive height anomalies are also frequently present during early onset. The study indicates that onset variability is partly forced by synoptic conditions, and the successful use of general circulation models to estimate onset will depend on their simulation of the zonally asymmetric component of the westerly circulation.

Corresponding author address: Dr. M. Tadross, Climate Systems Analysis Group, Department of Environmental and Geographical Science, University of Cape Town, Rondebosch 7701, South Africa. Email: mtadross@egs.uct.ac.za

1. Introduction

Rainfall in many parts of sub-Saharan Africa is a critical input to rain-fed agriculture and the production of staple grains. Over the Sahelian parts of West Africa, a long-term drying trend is observed (Nicholson 1994, 1995; Nicholson et al. 2000), with serious consequences for agriculture. Though a long-term drying trend is not noticed over southern Africa, both intraseasonal and interannual variability in summer (main cropping) season rainfall have been documented (e.g., Hulme 1992, 1996; Mason 1996; Usman and Reason 2004; Makarau and Jury 1997; Landman and Tennant 2000). Recent food security concerns within southern Africa arise in part from this variability and its impact on crop yield (Vogel 1994).

For many smallholder farmers who form the bulk of the rural population in southern Africa, crop yield depreciation is a matter of survival as farming is mainly for subsistence. Many of these farmers lack the capacity to adapt farming practices and are therefore vulnerable to climate variability. As Walter (1967) reported, a farmer’s main concern is for the rains to be consistent enough to guarantee sufficient soil moisture at planting and that those conditions are maintained or improved in the course of the season. This confirms the observation that variability in seasonal characteristics such as onset, cessation, and dry spell frequency are damaging to agriculture (Ati et al. 2002; Usman et al. 2004). Given information on the seasonal distribution of rainfall, a farmer can choose to plant a more or less drought-resistant crop (e.g., maize, sorghum, or millet), whereas information on the onset and cessation of the season enables the choice of a long- or short-season cultivar. The information need thus begins with the onset of the growing season, which has been identified as a priority forecast requirement for the agricultural sector within southern Africa (Usman et al. 2004; International Research Institute for Climate Prediction 2001). In the absence of information on cessation dates and dry spells, a successful onset forecast would allow farmers to assess the implications of an early/late start and importantly for some, given the prohibitive cost of seed for a second planting, warn against false starts to the season.

While the understanding of dry/wet spells has received some attention over southern Africa (e.g., Usman and Reason 2004; Makarau and Jury 1997; Matarira and Jury 1992), not much is known about the interannual variability in regionwide patterns in onset dates. The Famine and Early Warning System (FEWS) program of the United States Agency for International Development has been estimating the start of season (SOS) as part of a monitoring tool [Water Requirement Satisfaction Index (WRSI); information available online at http://edcintl.cr.usgs.gov/fews/wrsi.html] since 1996. This monitoring of the SOS is necessary to describe patterns and possible trends and resolve the associated implications for agriculture.

Previous studies have used many criteria for determining the onset of rains and this reflects both the available data and the purpose of the study. Chmielewski and Rötzer (2002) used phenology to assess variability in the start of season over Europe, whereas over the West African Sahel, the onset of the rains has been assessed and/or predicted using precipitation (e.g., Walter 1967; Kowal and Knabe 1972; Benoit 1977; Olaniran 1983; Stern et al. 1981; Sivakumar 1988; Omotosho et al. 2000; Sarria-Dodd and Jolliffe 2001; Ati et al. 2002), upper-atmospheric winds (e.g., Omotosho 1990, 1992, SST models (e.g., Lamb and Peppler 1992; Eltahir and Gong 1996; Janicot et al. 1998), and equivalent potential temperature (e.g., Usman 1999; Omotosho et al. 2000).

Our focus in this study is the growing of maize, the largest food staple in southern Africa (Smale and Jayne 2003). A rainfall-only definition and two independent datasets are used to determine the onset of the maize growing season. Characteristics of onset are compared in the two datasets, and the differences in pentad rainfall patterns during early and late onset are evaluated for regions of Zimbabwe and South Africa. Evidence for large-scale synoptic control of onset is presented, and the implications for agriculture are discussed.

2. Data and methods

a. Rainfall data

The following analysis utilizes the Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP), a merged satellite and rain gauge product (Xie and Arkin 1997). In this study, the data cover the period 1 January 1979 to 1 July 2002 at pentad time scales and at a spatial resolution of 2.5° × 2.5°. Though the grid resolution may be too coarse to capture localized effects due to vegetation and topography, for this analysis it is assumed that the dominant rain-producing systems in the region have large enough spatial extents to be well represented in the data. A comparison of this dataset with some of the underlying rain gauge data [communicated over the General Telecommunication System (GTS)] for the purpose of calculating the onset of the maize growing season is given in Tadross et al. (2003). The study concluded that compared to the GTS gauge data, the CMAP data gave similar patterns of onset mean and trend over South Africa and Zimbabwe, indicating that the use of satellite data did not significantly bias the estimation of onset.

A second more detailed dataset is also used to study the longer-term variability over South Africa and verify the CMAP data where the two are coincident. This station-only dataset was compiled by the Computing Center for Water Research (CCWR) of South Africa (Dent et al. 1989). It is made up of daily records covering South Africa at a spatial resolution of 0.5° for the period 1950–97. For the purposes of this study, the CCWR data were first averaged to the same pentad periods found in the CMAP data.

In the following analysis, the CCWR data are assumed to be the most reliable dataset. The reason for this assumption is demonstrated in Fig. 1, which shows the average number of stations contributing to the calculation of rainfall in each gridded bin for the CCWR (Fig. 1a) and the GTS (gridded to the CMAP spatial resolution; Fig. 1b) datasets. The GTS station data form a subset of a larger set of stations that contribute to the CMAP rainfall estimates. However, the spatial map of station density in Fig. 1b reflects the spatial distribution of random error estimated as part of the CMAP data production (Xie and Arkin 1997), with regions of high GTS station density associated with low CMAP random errors and vice versa. Given these observations and those noted in Tadross et al. (2003), the CMAP onset estimates are utilized only for South Africa and Zimbabwe. Figure 1 also highlights the difference in spatial resolution between the two datasets (CMAP is a factor of 5 coarser than CCWR) and the high station density in the CCWR data.

Also shown in Fig. 1b are the pixels representing the Free State (FS), southwest Zimbabwe (SZ), west Zimbabwe (WZ), and central Zimbabwe (CZ). All four pixels have a minimum average GTS station density of three, and time series of onset from these points are used later to define early and late onset.

b. Definition of onset

Both datasets were used to calculate the pentad at which the onset of the maize growing season occurred. The criteria for defining onset are those used by FEWS and given in AGRHYMET (1996). Onset is calculated as the amount of rain required to successfully germinate and grow maize during the first month after planting; the first “dekad” (10 days) must have a total rainfall of 25 mm, and this must be followed by two dekads with a total of at least 20 mm of rain. These conditions take care of the initial moisture requirement for seed germination and crop establishment and the need to avoid false starts by ensuring that soil moisture levels are high enough to sustain initial crop development.

Millet and sorghum are also grown in southern Africa, especially where rainfall variability is high such as in the WZ and SZ regions. Both crops have shorter growing periods and require less moisture than maize. However, like all cereals, sorghum and millet require consistent water during their germinating phase, which if not satisfied significantly decreases their yield (Fischer et al. 2004). Both crops also have less nutritional and market value than maize (Fischer et al. 2004), so farmers prefer and will risk growing maize if they think a good rainfall season is ahead. It is for these reasons that we believe that the variability in onset presented here is relevant to the WZ–SZ regions and to the cultivation of sorghum and millet.

The search for the pentad that marked onset, based on the above criteria, started at the pentad centered on 6 August each year, which is before the summer growing season. The onset was then calculated relative to this first pentad; for example, pentad 12 is the pentad following 2 October.

c. Pattern identification using self-organizing maps

A nonlinear pattern recognition technique, known as a self-organizing map (SOM), is used to identify rainfall systems (as represented by the pentad data) and 500-hPa geopotential height anomalies typically found during early and late onset. Although not a new technique, SOMs have only recently been introduced into the climate literature (e.g., Hewitson and Crane 2002; Tennant and Hewitson 2002). The technique seeks to identify a set of archetypes that represent a generalization of the data structure, and which span the data space. The SOM first randomly initializes a two-dimensional user-defined number of random patterns (n × m) to fit the data space. The (n × m) dimension is subjective and determines the degree of generalization. However, the SOM solution is robust under varying dimensions of m and n.

The SOM then undergoes a period of training where it adjusts each of these patterns (known as “nodes”) to span the data space. At the end of the training, each node represents a particular archetype of a local region within the data space. Each datum sample (each rainfall or geopotential height pattern) can be mapped to a particular node, effectively clustering the data. In this way, the frequency of occurrence of this general pattern can be assigned to each node. The resulting frequency surface (across the n × m node surface) represents the two-dimensional histogram of the generalized archetypes. The technique differs from traditional cluster analysis in two respects. First, the grouping of data samples is a postprocessing step, and the SOM algorithm is not explicitly clustering the data but rather seeks to identify archetypes that representatively span the data space. Second, the node array of identified archetypes is constructed as a generalized projection of the continuum nature of the data space. In this manner, adjacent nodes are correlated, while opposing corners of the array are opposition modes, with the two diagonals of the array analogous in some respects to the first two empirical orthogonal functions of the data. It is important to note that the nodes are not equidistant in data space. The SOM allocates more SOM nodes where there is more information content in the data space to warrant it. Detailed discussion of the SOM theory and possible applications can be found in Hewitson and Crane (2002).

The CMAP-derived onset time series for each of the four areas in Fig. 1b are used to separate years of early and late onset, the pentad rainfall patterns from which are subjected to an SOM in section 4. To gain a clearer insight into the atmospheric dynamics associated with late and early onset, we use an SOM to generalize daily eddy geopotential heights, calculated from the National Centers for Environmental Prediction (NCEP) reanalysis. This analysis is presented in section 5. Our convention throughout this paper will be to refer to each austral summer season by the year in which the season starts; for example, the summer season spanning the years 1984 and 1985 is the 1984 season.

3. Mean, standard deviation, and trends in onset

The mean onset pentad (after 3 August) for both the CCWR and CMAP datasets is presented in Fig. 2. It is shown for the period 1979–97, which was common to both datasets; differences using CCWR data for the 1950–97 period were mostly less than 1 pentad with an average change of +0.04 pentads over northeast South Africa. CMAP (Fig. 2a) indicates that onset is attained after mid-November (pentad number 20), if at all, in the climatologically dry southwestern regions. Late onset (late November into December) also occurs over northern Mozambique and western Madagascar. Over the climatologically wet regions—Angola to the north and South Africa to the south—mean onset occurs earlier in September or October. The rest of the subregion is characterized by a broad band of mean onset in late October and November.

Onset anomalies for each year (not shown) exhibit interannual variability; the region of later onset in the southwest sometimes expands to the northeast. Areas to the east never attain onset during certain seasons, for example, southeast Zimbabwe (1982 and 1991), southern Madagascar (1982), and southern Mozambique (1991 and 1997). These years are conspicuous by their association with El Niño events in the equatorial Pacific, and the strong association between southeastern Zimbabwe rainfall and ENSO has been noted in previous work (Matarira 1990b). CCWR mean onset is shown in Fig. 2b, and it largely reflects the pattern seen over South Africa in the CMAP data—the characteristic bulge of early onset to the southeast and the region of later onset to the west.

Figure 3 shows the standard deviation of onset for the common 1979–97 period. The CMAP data (Fig. 3a) indicates a northeast–southwest gradient over South Africa with higher variability to the west and lower variability to the east. This is in broad agreement with the CCWR data (Fig. 3b) though the CCWR data indicate higher variability to the east with a minimum standard deviation of 3 pentads (15 days). Farther north over Zimbabwe, the CMAP data indicate a standard deviation of approximately 3–5 pentads.

Of interest to farmers in the region is the notion that onset is becoming progressively later. A robust regression (Press et al. 1993) of onset, which is influenced less by outlying data points than a linear regression, is shown in Fig. 4. The trends are not significant as onset is highly variable (Fig. 3). However, in general the CMAP data (Fig. 4a) indicate that onset has on average been getting later during the 1979–97 period, with parts of South Africa and Zimbabwe changing at a rate of 1–2+ days yr−1. The CCWR data (Fig. 4b) again indicate insignificant trends, with onset in parts of the north and west trending toward earlier dates. These same regions exhibit the smallest trends in the CMAP data. Observable in both the CCWR and CMAP data are the high positive trends found over the southeast coast and the region around the Limpopo River (northern border of South Africa). The latter region is shown to be changing on average by more than 2.5 days yr−1 in places (47 days over the 19-yr period). This trend for later onset in the Limpopo province has been recorded in interviews made by the author with farmers who noted it as a cause for concern.

To establish the sensitivity of the results to the definition of onset, the above estimates of mean, standard deviation, and trend were repeated using an onset criterion of 45 mm of rain falling in a single dekad. This definition implies heavier rainfall than our original definition, and the most noticeable difference was that mean onset occurred later (same spatial pattern), with the standard deviation and trend of onset remaining similar in each dataset.

Accounting for differences in spatial resolution, Figs. 2 –4 demonstrate reasonable agreement between the CCWR and CMAP data over South Africa during the 1979–97 period. Recalculating these attributes of onset using the CCWR data for the 1950–97 period, the mean and standard deviation remained similar. The trends are reduced (Fig. 5; note the smaller scale), but still indicate a mean change of 24 days or more in places. Also indicated in Fig. 5 are the topographic contours, which show the steep escarpment to the southeast and the Limpopo valley to the north. Although not significant, the positive trends mostly follow these low-lying coastal regions and the Limpopo valley (also seen in Fig. 4). This indicates physical consistency in the trends and as demonstrated later, may be related to changes in synoptic systems that steer moisture inland from the neighboring ocean. Figure 6 further demonstrates decadal variability in onset (5-yr mean) over three subregions and northeast South Africa (28.5°–22.5°S, 25°–33°E). The late 1970s is a period when onset was generally earlier than the late 1990s. This explains the positive trends in Fig. 4 and the smaller trends in Fig. 5, suggesting decadal variability over South Africa and that onset was generally as late in the 1960s as it was in the 1990s.

Figures 2 –6 demonstrate that there has been significant interannual and decadal variability of onset over parts of South Africa and Zimbabwe. The following section examines the different pentad rainfall patterns in the CMAP data that are associated with this variability.

4. Rainfall systems associated with early and late onset

To more clearly assess the nature of the changes in onset and the implications for agriculture, the different pentad rainfall patterns associated with early and late onset are categorized using an SOM. Each pentad of August–January CMAP rainfall data (for the 23 seasons of 1979–2001) are used to train the SOM, by which the characteristic spatial precipitation patterns (at pentad time scales) across the domain are identified. Through subjective evaluation of different SOM array dimensions a 6 × 4 node array was chosen, giving a total of 24 archetypes. The dimensions of the SOM array do not change the results but determine the degree of generalization of the data. Given the relatively small number of pentad data, the χ2 significance tests are dependent on the number of degrees of freedom (number of nodes). The 6 × 4 array provides an appropriate generalization of archetypes, while retaining both good discrimination between the typical rainfall patterns and reasonable χ2 separation of early and late onset.

Figure 7 shows the SOM nodes, with rainfall mostly below 1 mm day−1 over southern Africa in the bottom left node (1, 1) and intense rainfall in the top right node (6, 4). This configuration was adopted, as there is a good distinction of synoptic systems known to be responsible for rainfall over the region, for example, the representation of tropical temperate troughs (TTTs) in the nodes found toward the middle top (3, 4 and 4, 4). These TTTs are known to be responsible for the majority of the rainfall variability during the summer months of November–March (Washington and Todd 1999; Todd and Washington 1999). The nodes toward the middle bottom of the SOM (3, 1 and 3, 2) indicate a “corridor” of light rainfall across Botswana and southern Zimbabwe, following the Limpopo valley. It separates regions of heavier rainfall to the north and south and coincides with the area identified by Usman and Reason (2004) as experiencing the highest dry spell frequency across the subcontinent. As noted previously, it is along this corridor that onset was not attained during recent El Niño events.

To examine differences in the distribution of rainfall systems that define the onset, 6 pentads immediately after and including the onset pentad were extracted for each of the four regions shown in Fig. 1b. For each region, the 23 (1979–2001) onset dates were categorized as being either early or late depending on whether the onset that year was more than two pentads earlier/later than the average onset in that region. If a year was classified as early, then the 6 pentads during onset were grouped with data from other years that were classified as early. This was repeated for similarly classified late onsets and for each region, resulting in the following number of years being classified as early/late: 5/10 (FS), 7/8 (SZ), 8/7 (WZ), 7/10 (CZ). Of these four early/late classified sets, SZ and WZ are the most similar with 5 early and 4 late years common to both regions. These eight datasets (early/late for each of the four regions) were mapped to the SOM shown in Fig. 7, and the resulting frequency distributions were used to create an anomaly of late minus early frequencies.

Figure 8 shows the anomaly frequency distributions for each region; that is, positive values indicate rainfall systems more prevalent during late onset, and negative values indicate those more prevalent during early onset. Taking each node as a sample of the two-dimensional histogram, a χ2 test was executed on the two distributions (late and early) for each geographical region. The probabilities (p) that the late and early distributions are similar are stated in each panel. All probabilities are <0.1, indicating that the distributions are significantly different at the 90% confidence level.

Rainfall during early onset is characterized by low-intensity rainfall systems as defined by nodes toward the left-hand side of the SOM. The higher-intensity rainfall in the nodes toward the right-hand side of the SOM characterize the rainfall during late onset. For example, the TTT pattern in node 4, 4 is more frequent during late onset over SZ, WZ, and CZ. This is not surprising given that late onset occurs later during the summer rainfall season when the thermal low and convective systems over the continent are more developed. However, rainfall of greater intensity may not be favorable for maize yield if it leads to waterlogging, especially early in the growth cycle (Lizaso and Ritchie 1997).

To examine the possibility that preseason rainfall could be used as an indicator of onset variability, the late-minus-early anomaly of August rainfall frequencies is indicated in Fig. 9. This follows the same procedure as outlined above except that the 6 pentads of data were taken from the August preceding onset. Again all four regions have low χ2 probabilities indicating different frequency distributions significant at the 93% level or higher. Late onset over both WZ and CZ (the two northernmost regions) is associated with nodes 1, 2 and 1, 3, whereas early onset is associated with node 1, 1. This asymmetry is reversed for FS (Fig. 9a). Node 1, 1 is notable in that it represents heavier rainfall over the northeastern coast of Madagascar and less rainfall over the continent, whereas nodes 1, 2 and 1, 3 do not indicate such an intense rainfall maximum. This observation will be discussed further in section 6.

The above results illustrate the differences in rainfall patterns that contribute to early and late onset over the four regions. The following section seeks a clearer understanding of the atmospheric dynamics by examining the associated midtropospheric circulation features.

5. Midtroposphere dynamics during early and late onset

To illustrate the different synoptic states associated with early and late onset, we examine the midtropospheric features frequently found in each case. An SOM is used to analyze NCEP 500-hPa geopotential heights that are coincident with the rainfall data presented in the previous section. Daily averaged anomalies from the zonal mean were calculated and a 4 × 3 SOM trained on these August–January 1979–2001 daily eddy geopotential anomalies. This SOM size was chosen because on visual inspection it provided a good distinction between the different patterns and as the following discussion suggests, the patterns are dynamically realistic.

Figure 10 indicates the resulting SOM in which it can be seen that the largest anomalies represent the passage of low and high pressure systems in the westerlies to the south of the continent. Patterns to the left side of the SOM mostly reflect a positive height anomaly to the southeast of the subcontinent. The opposite is true for the top-right corner of the SOM where the trough is to the east and the positive anomaly is to the west. In the middle of the SOM, the anomalies are generally weaker and not so distinct. The degree to which the patterns capture the progression of troughs in the westerlies south of the continent is illustrated by moving around the SOM nodes in an anticlockwise direction; the trough moves to the east and then reappears in the west again moving eastward as the SOM nodes are traversed.

Figure 11 shows the late-minus-early SOM frequency anomaly in the 30 days after onset (coincident with the rainfall in Fig. 8). All distributions satisfied a χ2 difference test at the 99% confidence level or higher. Figure 11 clearly shows that over all four regions, the nodes to the left of the SOM, which represent the highest positive anomalies to the south/southeast of the continent, are frequently present during early onset. These positive anomalies are associated with onshore flow in the lower troposphere, which advects moisture over the continent from the warm Agulhas Current and Mozambique Channel. A decline in the frequency of these patterns would result in less moisture available for rainfall, especially over the lower coastal regions and the Limpopo valley. This is suggested as the mechanism responsible for the positive trends in onset noted in section 3.

However, there are noteworthy differences between the late-minus-early frequencies over each region. For instance, over FS and CZ, node 4, 1 is frequently present during early onset. This node represents a high pressure anomaly to the east and a trough farther west with its leading edge over the continent. The placement of this trough is known to be conducive for rainfall over South Africa and surrounding regions (Tyson and Preston-Whyte 2000; Harrison 1984).

Late onset over all four regions is associated with the nodes 4, 2 and 3, 3 and the central patterns. The central patterns are characterized by weak anomalies, which may be indicative of the retreat of the westerlies southward as the austral summer progresses. The two other nodes represent weak positive anomalies and troughs in the region to the southeast of the continent. This is the same region where positive anomalies are associated with early onset, and the placement of these positive and negative anomalies is discussed further in the following section.

Figure 12 shows the late-minus-early SOM frequency anomaly of 500-hPa eddy geopotential patterns in the 30 days during the August preceding onset (coincident with the rainfall in Fig. 9). The differences between the late and early distributions are significant at the 99% confidence level. Figure 9 demonstrates a higher frequency of intense rainfall over northeast Madagascar when onset is early over CZ and WZ. Figure 12 indicates that height anomalies characterized by the four upper-left nodes in Fig. 10 are often present when there is a higher frequency of these heavy Madagascan rainfall events. These four nodes (1, 2; 2, 2; 1, 3; and 2, 3) represent positive height anomalies to the south/southeast of the continent, positioned closer to the continent than similar height anomalies represented by nodes 1, 1 and 2, 1. The implication is that both the strength and subtle changes in the position of these anomalies are important. The relationship of these anomalies to the observed rainfall is discussed presently.

6. Discussion

Given the dynamical differences between early and late onset, we now attempt an explanation based on this and previous work. Section 5 of this paper demonstrated that positive 500-hPa geopotential height anomalies to the east of the continent and south of Madagascar are frequently present during early onset. Depending on the position of these anomalies, they may affect the lower troposphere in one of two ways. First, if they are in a northerly position, they may extend the position (or increase the strength) of the south Indian Ocean anticyclone to the west, which then advects moisture from the western Indian Ocean over the continent. Makarau and Jury (1997) noted this as a feature of wet spells over Zimbabwe and, given the association implied by Figs. 9 and 12, this would seem a logical explanation for the heavier rainfall over northeast Madagascar during August when onset is early over WZ and CZ. Second, if the positive anomalies are farther south and west, they may be associated with high-pressure ridging from the Atlantic Ocean, which brings moisture to the coast from the south and east. These ridging events are also known to be associated with wet spells over Zimbabwe (Matarira 1990a). These two mechanisms explain how the flow of air and moisture from the surrounding oceans to the south and east could be steered onshore. Changes in these mechanisms will be most apparent in rainfall over the low-lying coastal regions and the Limpopo valley, which receive the majority of the moisture. The consistent trend for later onset in these regions (Fig. 5) may be due to long-term changes in these mechanisms. However, it is important to note that the rainfall during onset is not conditional based on the mechanisms we have suggested here. The positive geopotential anomalies are also present before onset (Fig. 12) and may precondition the moisture availability before onset, and local convective activity is then responsible for the rain during onset.

It was noted in section 3 that during the El Niño events of 1982/83 and 1991/92, onset was not attained over parts of southern Zimbabwe. This observation and the weak positive correlations of onset in this region with SSTs in the central Pacific (not shown), suggests that El Niño is associated with later onset over southern Zimbabwe. This association is dynamically consistent with results presented here and elsewhere. Cook (2001) used a simple GCM and perpetual December El Niño SSTs to show that during El Niño the western part of the south Indian Ocean anticyclone is weakened and enhanced storm activity is found east of South Africa. This result is further reinforced by the findings of Tennant (2002), who showed that the region to the southwest of the continent is a region of high eddy available potential energy, which is mostly converted to transient energy. During dry El Niño years, this region is displaced to the northeast, and the associated transients enter the region south of Madagascar. The 500-hPa eddy geopotentials associated with these transients take the form of a trough to the south of Madagascar, which precludes the high pressure anomaly that is associated with early onset. Matarira and Jury (1992) observed a similarly positioned trough as a circulation anomaly associated with midsummer dry spells over Zimbabwe, and both Camberlin et al. (2001) and Jury (1996) noted it as a seasonal feature of dry El Niño summers.

Other work also highlights changes in the westerly circulation as a source of interannual and decadal variability; Mo (2000) showed that the quasi-biennial component (wavenumber-3 pattern in the midlatitudes) of the Southern Hemisphere circulation was strongly negative during the mid-1950s and late 1970s, the same periods when average onset over northeast South Africa was early. The same study also revealed that this pattern was most highly correlated to SSTs during the September–November period (when onset occurs in the region). Recent work related to the Antarctic Circumpolar Wave (Venegas 2003) demonstrates that the Southern Hemisphere SST and sea level pressure fields can be decomposed into a wavenumber-2 (ENSO forced) and -3 (internal to the Southern Hemisphere mid- to high latitudes) component. The wavenumber-3 component exhibited a peak spectral density at a period of 3.3 yr and increased in strength from the mid-1970s through to 1998, similar to the trends seen in Mo (2000) and Fig. 6. These wavenumber-3 patterns of variability are also seen at intraseasonal time scales of 10–50 days (Kiladis and Mo 1998), which suggests that they are relevant to discussions of intraseasonal characteristics such as onset. The trends in Southern Hemisphere circulation anomalies seen in these studies, and those of onset demonstrated in Fig. 6, further emphasize the evidence presented in this paper for the westerly circulation as a source of synoptic control of onset variability.

7. Summary

We have demonstrated features of the interannual variability of the onset of the maize growing season over Zimbabwe and South Africa as a starting point in the effort to predict this important seasonal characteristic. Within the region, it has been identified as one of the most useful pieces of information that subsistence farmers require in their preparations for planting. It is therefore recommended as a focus for research to better understand the physical controls of its variability.

Onset has been demonstrated to be getting later over South Africa and Zimbabwe during the period 1979–2001, but over South Africa this was shown to be part of longer-term variability. Later onset is associated with heavier rainfall, which could have negative consequences for crop yield if it leads to waterlogging. Preonset (August) rainfall is shown to be a possible indicator of early onset over parts of Zimbabwe. Such an indicator may be useful for prediction; it could relate to preseason perturbations in moisture availability or it may indicate preferential synoptic conditions for rainfall.

During early onset, a higher frequency of positive midtropospheric height anomalies were found to the southeast of the continent. However, a clear distinction of the effect of these anomalies is beyond the scope of this study; they may advect moisture into the region before onset and create the conditions for onset rainfall, or the rainfall during onset may be a direct consequence of the associated circulations. It is likely a combination of both, and research is presently underway to ascertain the effects of local feedbacks using a regional climate model.

It is expected that local vegetation, orography, and soil moisture will all play a role and that onset will vary significantly on small spatial scales. Even so, the link to large-scale dynamic atmospheric anomalies suggests that a general circulation model may be used for prediction. If the relationship between the large-scale dynamics and onset is stationary, an empirical downscaling of the onset of the growing season from the model predictions may be feasible. However, the model will need to accurately capture the local dynamics of the Southern Hemisphere westerly circulation and those of the Indian Ocean subtropical high.

Acknowledgments

CMAP rainfall and NCEP–NCAR reanalysis data were kindly provided by the NOAA–CIRES Climate Diagnostics Center, Boulder, Colorado. We are grateful to Dr. Remi Tailleux, Dr. Emma Archer, and two anonymous referees for their helpful comments and suggestions. Dr. M. T. Usman would like to acknowledge the University of Cape Town for providing post-doctoral fellowship support.

REFERENCES

  • AGRHYMET, 1996: Méthodologie de suivi des zones à risque. AGRHYMET FLASH, Bulletin de Suivi de la Campagne Agricole au Sahel 0/96, Vol. 2., 2 pp. [Available from Centre Regional AGRHYMET, B.P. 11011, Niamey, Niger.].

  • Ati, F. O., , C. J. Stitger, , and E. O. Oladipo, 2002: A comparison of methods to determine the onset of the growing season in northern Nigeria. Int. J. Climatol., 22 , 731742.

    • Search Google Scholar
    • Export Citation
  • Benoit, P., 1977: The start of the growing season in northern Nigeria. Agric. Meteor., 18 , 9199.

  • Camberlin, P., , S. Janicot, , and I. Poccard, 2001: Seasonality and atmospheric dynamics of the teleconnection between African rainfall and tropical sea-surface temperature: Atlantic vs. ENSO. Int. J. Climatol., 21 , 9731005.

    • Search Google Scholar
    • Export Citation
  • Chmielewski, F. M., , and T. Rötzer, 2002: Annual and spatial variability of the beginning of growing season in Europe in relation to air temperature changes. Climate Res., 19 , 257264.

    • Search Google Scholar
    • Export Citation
  • Cook, K. H., 2001: A Southern Hemisphere wave response to ENSO with implications for southern Africa precipitation. J. Atmos. Sci., 58 , 21462162.

    • Search Google Scholar
    • Export Citation
  • Dent, M. C., , S. D. Lynch, , and R. E. Schulze, 1989: Mapping mean annual and other rainfall statistics over southern Africa. Water Research Commission Tech. Rep. 109/1/89, Pretoria, South Africa, 230 pp.

  • Eltahir, E., , and C. Gong, 1996: Dynamics of wet and dry years in West Africa. J. Climate, 9 , 10301040.

  • Fischer, G., , H. van Velthuizen, , F. Nachtergaele, , and S. Medow, cited. 2004: Global agro-ecological zones—2000. Food and Agriculture Organisation of the United Nations and International Institute for Applied Systems Analysis. [Available online at http://www.fao.org/WAICENT/FAOINFO/AGRICULT/AGL/agll/gaez/index.htm.].

  • Harrison, M. S. J., 1984: A generalized classification of South African summer rain-bearing synoptic systems. J. Climatol., 4 , 547560.

    • Search Google Scholar
    • Export Citation
  • Hewitson, B. C., , and R. G. Crane, 2002: Self Organizing Maps: Applications to synoptic climatology. Climate Res., 22 , 1326.

  • Hulme, M., 1992: Rainfall changes in Africa: 1931–1960 and 1961–1990. Int. J. Climatol., 12 , 685699.

  • Hulme, M., 1996: Recent climatic change in the world’s drylands. Geophys. Res. Lett., 23 , 6164.

  • International Research Institute for Climate Prediction, 2001: Proceedings, Communication of Climate Forecast Information Workshop. International Research Institute for Climate Prediction, 51 pp.

  • Janicot, S., , A. Harzallah, , B. Fontaine, , and V. Moron, 1998: West African monsoon dynamics and eastern equatorial Atlantic and Pacific SST anomalies (1970–88). J. Climate, 11 , 18741882.

    • Search Google Scholar
    • Export Citation
  • Jury, M. R., 1996: Regional teleconnection patterns associated with summer rainfall over South Africa, Namibia and Zimbabwe. Int. J. Climatol., 16 , 135153.

    • Search Google Scholar
    • Export Citation
  • Kiladis, G. N., , and K. C. Mo, 1999: Interannual and intraseasonal variability in the Southern Hemisphere. Meteorology of the Southern Hemisphere, Meteor. Monogr., No. 49, Amer. Meteor. Soc., 307–336.

  • Kowal, J. M., , and D. T. Knabe, 1972: An Agroclimatological Atlas of the Northern States of Nigeria. Ahmadu Bello University Press, 15 pp.

    • Search Google Scholar
    • Export Citation
  • Lamb, P. J., , and R. A. Peppler, 1992: Further case studies of tropical Atlantic surface atmospheric and oceanic patterns associated with sub-Saharan drought. J. Climate, 5 , 476488.

    • Search Google Scholar
    • Export Citation
  • Landman, W. A., , and W. J. Tennant, 2000: Statistical downscaling of monthly forecasts. Int. J. Climatol., 20 , 15211532.

  • Lizaso, J. I., , and J. T. Ritchie, 1997: Maize shoot and root response to root zone saturation during vegetative growth. Agron. J., 89 , 125134.

    • Search Google Scholar
    • Export Citation
  • Makarau, A., , and M. Jury, 1997: Seasonal cycle of convective spells over southern Africa during austral summer. Int. J. Climatol., 17 , 13171332.

    • Search Google Scholar
    • Export Citation
  • Mason, S. J., 1996: Rainfall trends over the Lowveld of South Africa. Climatic Change, 32 , 3554.

  • Matarira, C. H., 1990a: Frequency and tracks of anticyclones and their effect on rainfall patterns over Zimbabwe. Theor. Appl. Climatol., 42 , 5366.

    • Search Google Scholar
    • Export Citation
  • Matarira, C. H., 1990b: Drought over Zimbabwe in a regional and global context. Int. J. Climatol., 10 , 609625.

  • Matarira, C. H., , and M. R. Jury, 1992: Contrasting meteorological structure of intra-seasonal wet and dry spells in Zimbabwe. Int. J. Climatol., 12 , 165176.

    • Search Google Scholar
    • Export Citation
  • Mo, K. C., 2000: Relationships between low-frequency variability in the Southern Hemisphere and sea surface temperature anomalies. J. Climate, 13 , 35993610.

    • Search Google Scholar
    • Export Citation
  • Nicholson, S. E., 1994: Recent rainfall fluctuations in Africa and their relationships to past conditions over the continent. Holocene, 4 , 121131.

    • Search Google Scholar
    • Export Citation
  • Nicholson, S. E., 1995: Variability of African rainfall on interannual and decadal time scales. Natural Climate Variability on Decade-to-Century Time Scales, Climate Research Committee SE, National Academy Press, 32–43.

    • Search Google Scholar
    • Export Citation
  • Nicholson, S. E., , B. Some, , and B. Kone, 2000: An analysis of recent rainfall conditions in West Africa, including the rainy seasons of the 1997 El Niño and the 1998 La Niña years. J. Climate, 13 , 26282640.

    • Search Google Scholar
    • Export Citation
  • Olaniran, O. J., 1983: The onset of the rains and the start of the growing season in Nigeria. Niger. Geogr. J., 26 , 8188.

  • Omotosho, J. B., 1990: Onset of thunderstorms and precipitation over northern Nigeria. Int. J. Climatol., 10 , 849860.

  • Omotosho, J. B., 1992: Long-range prediction of the onset and end of the rainy season in the West African Sahel. Int. J. Climatol., 12 , 369382.

    • Search Google Scholar
    • Export Citation
  • Omotosho, J. B., , A. A. Balogun, , and K. Ogunjobi, 2000: Predicting monthly and seasonal rainfall, onset and cessation of the rainy season in West Africa using only surface data. Int. J. Climatol., 20 , 865880.

    • Search Google Scholar
    • Export Citation
  • Press, W. H., , B. P. Flannery, , S. A. Teukolsky, , and W. T. Vetterling, 1993: Numerical Recipes in C: The Art of Scientific Computing. Cambridge University Press, 994 pp.

    • Search Google Scholar
    • Export Citation
  • Sarria-Dodd, D. E., , and I. T. Jolliffe, 2001: Early detection of the wet season in semiarid tropical climates of western Africa. Int. J. Climatol., 21 , 12511262.

    • Search Google Scholar
    • Export Citation
  • Sivakumar, M. V. K., 1988: Predicting rainy season potential from the onset of rains in the southern Sahelian and Sudanic climatic zones of West Africa. Agric. For. Meteor., 42 , 295305.

    • Search Google Scholar
    • Export Citation
  • Smale, M., , and T. S. Jayne, 2003: Maize in Eastern and Southern Africa: “Seeds” of success in retrospect. EPTD discussion paper 97, International Food Policy Research Institute, 79 pp.

  • Stern, R. D., , M. D. Dennett, , and D. J. Garbutt, 1981: The start of the rains in West Africa. J. Climatol., 1 , 5968.

  • Tadross, M., , B. C. Hewitson, , and M. Usman, 2003: Calculating the onset of the maize growing season over southern Africa using GTS and CMAP data. CLIVAR Exchanges, No. 27, International CLIVAR Project Office, Southampton, United Kingdom, 48–50.

  • Tennant, W. J., 2002: Event characteristics of intra-seasonal climate circulations. Ph.D. thesis, University of Cape Town, 185 pp.

  • Tennant, W. J., , and B. C. Hewitson, 2002: Intra-seasonal rainfall characteristics and their importance to the seasonal prediction problem. Int. J. Climatol., 22 , 10331048.

    • Search Google Scholar
    • Export Citation
  • Todd, M. C., , and R. Washington, 1999: Circulation anomalies associated with tropical-temperate troughs over southern Africa and the southwest Indian Ocean. Climate Dyn., 15 , 937951.

    • Search Google Scholar
    • Export Citation
  • Tyson, P. D., , and R. A. Preston-Whyte, 2000: The Weather and Climate of Southern Africa. 2d ed. Oxford University Press, 396 pp.

  • Usman, M. T., 1999: A convective instability index for summer monsoon onset determination over the Sahel. Proc. Workshop on the West African Monsoon Variability and Predictability (WAMAP), Dakar, Senegal, WMO.

  • Usman, M. T., , and C. J. C. Reason, 2004: Dry spell frequencies and their variability over southern Africa. Climate Res., 26 , 199211.

  • Usman, M. T., , E. Archer, , P. Johnston, , and M. Tadross, 2004: A conceptual framework for enhancing the utility of rainfall hazard forecasts for agriculture in marginal environments. Nat. Hazards, 34 , 111129.

    • Search Google Scholar
    • Export Citation
  • Venegas, S. A., 2003: The Antarctic Circumpolar Wave: A combination of two signals? J. Climate, 16 , 25092525.

  • Vogel, C. H., 1994: (Mis)management of droughts in South Africa: Past, present, future. South Afr. J. Sci., 86 , 382386.

  • Walter, M. W., 1967: Length of the rainy season in Nigeria. Niger. Geogr. J., 10 , 123128.

  • Washington, R., , and M. Todd, 1999: Tropical-temperate links in southern African and southwest Indian Ocean satellite-derived daily rainfall. Int. J. Climatol., 19 , 16011616.

    • Search Google Scholar
    • Export Citation
  • Xie, P., , and P. A. Arkin, 1997: Global precipitation: A 17-year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs. Bull. Amer. Meteor. Soc., 78 , 25392558.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Mean number of stations contributing to the rainfall in each gridded bin: (a) CCWR (1950–97) and (b) GTS (1979–2001); with an average of 3 to 5 stations per region.

Citation: Journal of Climate 18, 16; 10.1175/JCLI3423.1

Fig. 2.
Fig. 2.

Mean onset of the maize growing season (number of pentads after 3 Aug) for the 1979–97 period: (a) CMAP and (b) CCWR.

Citation: Journal of Climate 18, 16; 10.1175/JCLI3423.1

Fig. 3.
Fig. 3.

Std dev of onset (number of pentads) for the period 1979–97: (a) CMAP and (b) CCWR.

Citation: Journal of Climate 18, 16; 10.1175/JCLI3423.1

Fig. 4.
Fig. 4.

Trend in onset (pentads yr−1) for the period 1979–97: (a) CMAP and (b) CCWR.

Citation: Journal of Climate 18, 16; 10.1175/JCLI3423.1

Fig. 5.
Fig. 5.

CCWR trend in onset (pentads yr−1) for the 1950–97 period. Contours indicate height of local topography (m).

Citation: Journal of Climate 18, 16; 10.1175/JCLI3423.1

Fig. 6.
Fig. 6.

Time series of onset for the Limpopo, FS, Kwazulu–Natal (KZN), and northeast (28.5°–22.5°S, 25°–33°E) regions of South Africa.

Citation: Journal of Climate 18, 16; 10.1175/JCLI3423.1

Fig. 7.
Fig. 7.

Archetype pentad rainfall patterns associated with each node of the 6 × 4 SOM. Scale is in mm day−1.

Citation: Journal of Climate 18, 16; 10.1175/JCLI3423.1

Fig. 8.
Fig. 8.

Frequency anomaly of late-minus-early distribution of rainfall patterns (SOM nodes) in the six pentads during onset for four regions: (a) FS, (b) SZ, (c) CZ, and (d) WZ. Probabilities (p) are that the late and early frequency distributions of rainfall patterns are similar according to a χ2 test.

Citation: Journal of Climate 18, 16; 10.1175/JCLI3423.1

Fig. 9.
Fig. 9.

Same is in Fig. 8, but during Aug for the four regions.

Citation: Journal of Climate 18, 16; 10.1175/JCLI3423.1

Fig. 10.
Fig. 10.

Daily 500-hPa eddy geopotential height anomalies (m) associated with each node of the 4 × 3 SOM.

Citation: Journal of Climate 18, 16; 10.1175/JCLI3423.1

Fig. 11.
Fig. 11.

Frequency anomaly of late-minus-early distribution of daily 500-hPa eddy geopotential patterns (SOM nodes) in the 30 days during onset for four regions: (a) FS, (b) SZ, (c) CZ, and (d) WZ. Late and early frequency distributions are different at the 99% confidence level.

Citation: Journal of Climate 18, 16; 10.1175/JCLI3423.1

Fig. 12.
Fig. 12.

Same as in Fig. 11, but during Aug.

Citation: Journal of Climate 18, 16; 10.1175/JCLI3423.1

Save