Seasonal rainfall variability in Guinea Savanna part of Nigeria: a GIS approach Open Access

Rainfall regime is the most important climate factor influencing agricultural activities. Rainfall can vary considerably even within a few kilometres distance and on different time scale. This implies that crop yield is exceedingly variable over space and time. It has biggest effect in determining the crops that can be grown, the farming system, the sequence and timing of farming operations (Adejuwon, 2005). Rainfall can also be seen as the supplier of soil moisture for crops. The soil moisture supply, however, does not depend on rainfall alone, but also on various other factors concerned in the hydrological equation, such as evapotranspiration and surface run‐off (IPCC, 2000). In many areas with alternating wet and dry seasons the annual rainfall is less than the amount of the water that a crop well supplied with water would transpire during the growing season. Theoretically, there are three different forms of rainfall variability:

• 1.

  spatiotemporal;

• 2.

  inter‐annual; and

• 3.

  intra‐annual variability (NEST, 2003).

Spatiotemporal variability has to do with differences in total rainfall received between places structurally located within a given region over a period of time. Inter‐annual rainfall variation can be defined as the annual deviation from long‐term averages or the differences in rainfall between years. Intra‐annual rainfall variability refers to the distribution of rainfall within a year (Obasi, 2003a, b). In the last decade, inter‐annual rainfall variations are causes of great stress to the farming activities, crop production and crop yield in the Guinea Savanna of Nigeria (Adejuwon, 2004).

Previous studies have examined inter‐annual rainfall variability in Nigeria and other part of West Africa (Awosika et al., 1994; FAO, 2001; Obasi, 2003a, b; Adejuwon, 2004). It affects the various aspects of plant growth and yields; consequently, alter crop productivity. As observed by Awosika et al. (1994), the aggregate impact of drought on the economy of Nigeria in 1992 was between 4 and 6 percent of the gross domestic product (GDP). From an analysis of recent rainfall conditions in West Africa, FAO (2001) concluded that a long‐term change in rainfall has occurred in the semi‐arid and sub‐humid zones of West Africa. Although, it may appear that little or nothing could be done to improve variability in rainfall since most of its causes are natural. Thus, there is need for in‐depth study and understanding of spatiotemporal rainfall variability.

Surprisingly, little systematic research has focused on the distribution patterns of the impacts of rainfall variability in terms of mapping its spatiotemporal impact using the modern geographical information systems (GIS) techniques such as kriging interpolation technique. There is need however for an integrated this type of GIS modeling system, to allow agricultural producer as well as policy makers to know the impact of spatial‐temporal variation in rainfall on crop yield for better management, productivity and profitability.

Therefore, this paper aimed at using GIS kriging interpolation technique to examine and map the spatiotemporal variation in rainfall. Kriging technique is employed for this research work because data in respect of rainfall variability is often collected in respect of widely scattered points known as meteorological stations. One major problem is how to estimate values for the other locations in respect of which the primary data are not available. Scholars have been using climate models such as general circulation models (GCM) and this are capable of generating primary data in respect of the nodes of grids of widely spaced longitudes and latitudes (Adejuwon, 2005). Problems also often arise as to what values to attach to weather or climatic variables at points other than these nodes. There are now available standard techniques for solving these problems. Where the data have been collected at normal meteorological stations, the procedures are described as kriging interpolation technique. There is no doubt that an in‐depth understanding of rainfall variability as well as a broad knowledge of its spatiotemporal pattern using GIS would be very useful in the monitoring strategies which should improve the crop yield in Nigeria.

The study area is Guinea Savanna Ecological zone of Nigeria. The area lies between the semi‐arid north and the wet southern part of the country. The area lies approximately between longitudes 3° and 14°E and latitudes 7° and 10°N (Figure 1). The region is sometimes referred to as the middle belt and has a land extent of about 323,569 square kilometers and a west‐east breadth of about 800 kilometers (Buchanan and Pugh, 1955). The estimated total population of the entire study area was put at 88,257,112 with male totaling 53,623,124 and female 34,633,988, according to National Population Commission (2006) reports. Rainfall in the zone is largely seasonal and highly variable from year to year, with mean annual rainfall of between 1,500 and 1,800 millimeters in north and south, respectively. The delineation of the Guinea Savanna zone is based on the mean annual rainfall as well as the severity of the dry season. The southern limit of Guinea Savanna zone is based on a mean annual rainfall of at least 1,600 millimeters and lowest mean monthly relative humidity of not less than 70 percent. The northern limit of Guinea Savanna zone is at approximately 1,550 millimeters mean annual rainfall, with the lowest mean monthly relatively humidity at about 29 percent. By April or May the rainy season is underway in most areas south of the Niger and Benue river valleys. Farther north of Guinea Savanna, it is usually June or July before the rains really commence. The peak of the rainy season occurs in August, when air from the Atlantic covers the entire country. From September through November, the northeast trade winds generally bring a season of clear skies and lower humidity for this area.

From December to February, however, the northeast trade winds blow strongly and often bring with them a load of fine dust from the Sahara. These dust‐laden winds, locally known as the harmattan, often appear as a dense fog and cover everything with a layer of fine particles. The harmattan is more common in the Guinean Savanna but affects the entire country (Anuforom, 2004). The dry season extends over a period of about six to seven months, from October/November to March/April, while remaining months are wet. The climate of the study area is characterized with relatively high temperatures throughout the year. The average annual maximum varies from 35 to 31°C throughout the year while the average annual minimum is between 23 and 20°C. On the Jos Plateau and the eastern highlands, altitude makes for relatively lower temperatures, with the maximum of 28°C and minimum of 14°C. Frequent drought periods have been among the most notable aspects of Guinean Savanna climatic attributes in recent years. Climatologists regard the twentieth century as one of the driest periods within the last several centuries. The well publicized droughts of the 1970s and 1980s were only the latest of several significant such episodes to affect West Africa (FAO, 2001). At least two of these droughts have severely affected large areas of the Guinea Savanna Ecological zone.

Rainfall data for 1970‐2000 were collected from the archives of the Nigerian Meteorological Services, Oshodi Lagos. The rainfall data used in this study consist of monthly and annual rainfall total for the period 1970‐2000. The data are for ten stations located in the study area. The rainfall synoptic stations selected to represent the Guinea Savanna Ecological zone of Nigeria and for which the data were collected include: Enugu, Ilorin, Jalingo, Lokoja, Makurdi, Minna, Mokwa, Ogoja, Shaki, and Yola (Table I and Figure 1). The rainfall data used in this study were collected by the Nigerian Meteorological Services, using Dines Tilting Rainguage and the British Standard Rainguage.

Three spatial interpolation methods were chosen for this research work: inverse distance weighting (IDW) method and the spline (completely regularized) as the determinist methods; and ordinary kriging as the stochastic method. In order to analyze the interpolation quality, an evaluation by cross validation has been carried out. The cross validation started by eliminating one of the rainfall stations. The three different spatial interpolation methods are then applied to estimate the missing value on the basis of the remaining observed ones. This process has been carried out on each rain station. For each interpolation test on the three methods, the rainfall observed values Z(x) have been considered, the estimated values of Ž(x), and the errors e(x)=Z(x)−Ž(x). The mean of the errors and its standard deviation was then calculated for each interpolation method. IDW method and the spline are the less efficient interpolation methods for this research work. On the contrary, ordinary kriging is the method that allows the sharpest interpolation rainfall data and is the most representative.

Ordinary kriging was used to interpolate the point observations from a network of rainfall base stations. The software used for all interpolations was ArcGIS, produced by Environmental Systems Research Institute. Though, several kriging interpolation methods have been tested for the production of maps in this study: cokriging; universal kriging; residual kriging, and ordinary kriging but, after many attempts and qualitative and quantitative verifications, the last of theses – ordinary kriging – was chosen for the map productions. Ordinary kriging estimates for a rainfall and crop yield distributed variable at any unmonitored location are computed as a weighted average of the known values from a surrounding set of sampled points. Kriging weights are derived from a statistical model of spatial correlation expressed as semivariograms that characterize the spatial dependency and structure in the data. A major strength of the method is that measured spatial dependence in the weather parameter of interest.

Over the last decade, researches have greatly increased into the use of GIS in a variety of applications that involve the processing of climatological and meteorological data. Recent efforts using GIS for applications in climate and meteorology include traditional GIS strategies for hydromet database analysis and management, and the fusion of hydromet data with traditional GIS applications (Tveito et al., 2000; Tveito and Schöoner, 2002; Daly et al., 2004; Zbigniew and Danuta, 2005). But new techniques are emerging which exploit GIS features and capabilities to support the analysis of hydromet data in the GIS operating environment. GIS have now become a very important and widespread tool serving a variety of functions in many environmental sciences including meteorology and climatology (Tveito et al., 2000).

The statistical method that was employed for this study is coefficient of variation (CV). CV was used to determine rainfall variability. Although, there are many measures of variability, the two most widely used are the relative variability and the CV. CV measurement is the more efficient for this research work (Ayoade, 1988). Therefore, CV was used to examine rainfall variability and to assess the sensitivity or response (i.e. to assess impact) of crop yield to variability in rainfall. It is a measure of dispersion given by: Equation 1 The mean and the standard deviations of rainfall for each station are first calculated, and then the CV is determined as a percentage of the mean. The CV is mathematically expressed as: Equation 2 where δ, the standard definition is defined by: Equation 3 where R_f is the annual rainfall for a given period and R¯_f is the average annual rainfall.

In this paper, two phases of GIS database management were adopted. First, database design phase were executed and the second phase is implementation phase (Kufoniyi, 1995). GIS database design phase (i.e. data modeling) for this study was divided into four levels namely (Figure 2): reality phase, conceptual design phase, logical design phase, and physical design phase.

Reality phase is all about observing the study area as it naturally exist. These include all aspects that may or may not be previewed by man. This involves mental abstraction of the reality for a particular application or group of applications which guides the user's request. Generally, the reality for this paper is the study area (Figure 2) and this was viewed in terms of the rainfall variability, which is the basic requirement of this paper. Conceptual design phase involves the choosing of a data model (Kufoniyi, 1995) that includes: vector database model; tessellation database model; and object oriented database model. Logical design phase involves the representing the database model designed to reflect the recording of the data in the GIS. This process is also known as data structuring (Kufoniyi, 1998). Relational data structure was adopted for this study (Figure 2). Although, there are many other types of data structures like hierarchies; object‐oriented, object‐relational, among others, but relational data structure presents data in a simple uniform manner in form of tables or relations. Physical design phase involves choosing GIS subsystems that were used for the study (Figure 2). The GIS hardware components that were used for this research include digitizer, scanner, printers, among others. A computer system of Pentium IV Processor, 150GB, was used for the analysis and cartographic works. ILWIS 3.2 and Arc‐view GIS were used as GIS software for spatial analysis, interpolation and cartographic presentations.

Figures 3‐6 show variation in rainfall for different decade between 1970 and 2000, in Guinea Savanna part of Nigeria. The interpolation silhouette graduated from the Southern to Northern Guinea Savanna and the pattern is typical to Figures 3‐6. The decadal interpolation values vary from 550 to 2,987 millimeters. Highest interpolation silhouette was shown during the third decade, i.e. 1990‐1999 (Figure 3) with six interpolation silhouettes while the least was observed during the second decade, i.e. 1980‐1989 (Figure 4) with four interpolation silhouettes. It is observed that variation in rainfall is very high in most of Northern Guinea Savanna; Yola, Minna, Abuja, and Mokuwa with value of coefficient variation between 26 and 49 percent (Table II). In Southern Guinea Savanna, however, the CV is very low especially, in Ogoja, Enugu, and Shaki. This actually mean that the lower the inter‐annual rainfall total, the higher the value of CV which implies great variability in rainfall (Figures 6 and 7). The ten‐year mean rainfall was highest in the period 1970‐1979 (Figure 3) and least in 1980‐1999 (Figure 4). More so, the variation is very significant in Makurdi, Ilorin, and Yola (Table II) with lower rainfall values for the period of 1980‐1989 (Figure 4). Generally, rainfall value was increased in 1990s (Figure 5) for all the stations, which signified the peak. In Shaki, Lokoga, Ogoja, and Minna rainfall total is observed to be very low within the period of 1980s (Figure 4), and early 1990s (especially in 1991‐1993). The figure also shows abnormal low rainfall in 1980s with amount less than 900 millimeters in almost all the stations. Further more, there is increase in Mokurdi within the period of late 1990s (especially 1996‐2000), which reveals an increase with amount ranging between 1,200 and 2,200 millimeters Per annum, except in 1980s, which has less than 900 millimeters Per annum. In Mokwa, Shaki, Ogoja, and Minna, Figure 3 also shows that between 1980 and 1989 there is low rainfall of below 700 millimeters Per annum while from 1990 to 2000 (Figure 6) rainfall value is as high as 1,800 millimeters Per annum. The evidence of rainfall variability is observed as a result of the impacts of climate change and variability.

Generally, in Southern Guinea Savanna, the CV is very low especially in Enugu (9 percent) and Shaki (8 percent) (Table II). Meanwhile, the stations in the Northern Guinea Savanna have high coefficient at variation. For instance, Minna has 49 percent, Ilorin (43 percent), and Yola has 44 percent (Table II). This implies that the lower the inter‐annual rainfall total, the higher the value of CV and thus variability in rainfall. This actually implies that that rainfall really varies both in time and space. The impacts of its variability are very momentous on crop yield especially in the Guinea Savanna part of Nigeria. This may be as a result of evaporation potential that is very high throughout the year in the region. This paper also established that the rainfall variability generally increases with decreasing total rainfall (Ojo and Oni, 2001). Also, rainfall varies inversely with the mean rainfall and also rain becomes less reliable as one move towards the Northern Guinea Savanna of Nigeria.

Although both Southern and Northern Guinea Savanna are similar in that, they experience alternate wet and dry seasons of varying intensity at times of high and low radiation, respectively, but, it is not very satisfactory from an agricultural point of view to lump them all together. Apart from local differences due to differences in altitude (e.g. high altitude of Jos), there are remarkable variations in the number of rainfall per annum, in the duration of these seasons and in the amount of annual and seasonal rainfall (Figure 7). This confirms the hypothesis that rainfall varies greatly within the study area. Accordingly, variability of rainfall is an important factor in Guinean Savanna part of Nigeria where rainfall tends to be more seasonal in its incidence within the year.

Observation in Guinea Savanna shows that rainfall variability continues to be on the increase following the increase in climate change and variability as revealed in the analysis. The total rain and distribution of rains at any location determine the frequency and intensity of drought and flooding as well as the length of growing season in that location (Anuforom, 2004). The evidence of droughts and floods are observed as a result of the impacts of climate variability. The impacts of its variability are very significant on crop yield especially in Guinean Savanna Ecological zone of Nigeria where evaporation potential is very high throughout the year.

This paper also established that the rainfall variability generally increases with decreasing total rainfall (Ojo and Oni, 2001). It should be noted that rainfall varies inversely with the mean rainfall or that rain become less reliable as one moves toward thee Northern Guinean Savanna of Nigeria. Although, both southern and northern Guinean Savanna are similar in that they experience alternate wet and dry seasons of varying intensity at times of high and low sun, respectively, it is not very satisfactory from an agricultural point of view to lump them all together. Apart from local differences due to altitude, and presences of small lakes, there are important variations in the number of wet and dry seasons per annum, in the duration of these seasons and in the amount of annual and seasonal rainfall.

This confirms the hypothesis that rainfall varies greatly within the study area. The variability of rainfall is an important factor in Guinean Savanna of Nigeria where rainfall tends to be more seasonal in its incidence within the year. Also, the result of CV revealed that rainfall variability is very high in most of Northern Guinea Savanna (e.g. Yola, Minna, and Ilorin) with values of CV between 26 and 49 percent while in Southern Guinea Savanna, the CV is very low especially, in Enugu (9 percent) and Shaki (8 percent). Variability in the onset and the amount of rainfall is known to create food insecurity. In the years of drought in the Guinea Savanna zone in the 1980s, harvest failure was remarkable throughout the region. The situation led to the intensification of crop irrigation as a response mechanism. Close to one million livestock were lost, affecting meat and dairy supply throughout the country. Flood hazards in both the north and south of the region consistently posed a danger to farmlands. This phenomenon challenges the age‐long ability of farmers to predict when to plant their crops. Unpredictable changes in the onset of rains in the last 20 to 30 years have led to situations where crops planted with the arrival of early rains get smothered in the soil by an unexpected dry spell that can follow early planting. That, coupled with the late arrival of rains due to climate variability, results in harvest failures in the different ecosystems where rainfed agriculture is practiced.

Results from the study reveal that GIS techniques are exceedingly helpful in assessing rainfall variability. It is easy to calculate and map the mean monthly or seasonal rainfall for specified periods and the deviations from the mean value. It is hard to imagine contemporary rainfall monitoring without GIS applications. GIS tools give very exact and detailed images of analyzed data more effectively than traditional – usually manual techniques. GIS tools also enable the uncomplicated calculation and display of the area under specified rainfall conditions and the display of maps for rainfall assessment purposes.

Though, several spatial interpolation methods have been tested for the production of maps in this study like co‐kriging; universal kriging; residual kriging, and ordinary kriging. After many attempts of qualitative and quantitative verifications, the last of these – ordinary kriging – was chosen for the map productions. Thus, ordinary kriging interpolation method is found useful and others are less efficient interpolation methods for this research work. Ordinary kriging is the method that allows the sharpest interpolation rainfall data and is the most representative. Rainfall spatial interpolation methods range from simple estimations to complex procedures and can be gathered in many typologies. The one that are generally used by scholars is based on two categories; the determinist methods and the stochastic methods. The determinist methods gather the bary center techniques with IDW method for example, the area division techniques like the Thiessen polygons method, and the spline techniques (regularized and tension). The stochastic methods include the classic regression techniques with amongst them the trend surface analysis technic, the local regression techniques, and the kriging techniques (Burrough, 1990). Most interpolation techniques give similar results when data are abundant. For sparse data, the underlying assumptions about the variation among sampled points differ and, therefore, the choice of interpolation method and parameters become critical. Kriging has the disadvantage of high‐computational requirements (Burrough and McDonnell, 1998). However, several studies conclude that the best quantitative and accurate results are obtained by kriging (Burrough and McDonnell, 1998).

Crop production has been the most essential sector of the economy in Nigeria, accounting for more than 50 percent of GDP. In order to minimize the negative impact of climate change/variability (consequential of variation in rainfall), a number of adaptation measures are open to Nigeria. There is no doubt that farmers have the potential to adapt to rainfall variability not only by physiological, but also by social and cultural adaptive measures, such as finding new crop species that would have resistance to fluctuation and reduction in rainfall and agricultural traditions. Owing to these abilities of adaptation, farmers can cultivate throughout the year.

Therefore, it is necessary to study the capacity for adaptation to extreme rainfall events. Adaptation measure include encourage researches in finding new crop species and varieties that would have resistance to fluctuation and reduction in rainfall. It is very important that hybrids of crops that are well acclimated as well as drought and pest resistant be developed. Meteorological agency need to constantly alert farmers on weather. Farmer should also be encourages to use meteorological forecasts. New irrigation schemes should be introduced to dry land management in Nigeria, which will improve water use efficiency and minimize moisture stress for crops. This is particularly relevant in Guinean Savanna zone where climate variability is expected to result in reduced amount of rainfall for rainfed agriculture. Agroclimatological system should be developed and implemented for accumulation and efficient use of rainfall.

Farmers' responses to seasonal variation in rainfall may take many other forms. Adaptation to extreme rainfall variability may include agro‐pastoral management techniques providing for a more efficient use of reduced rainfall. Poverty and hunger resulting from reduction in rainfall may cause migration and degradation, or change of diet. Agricultural production could be increased by doubling the crop areas or by investing in agriculture management and technology. Producing genetically drought‐resistant crops would help, as would better water resource management, more efficient storage systems, improved processing methods, better pest management. A number of government policies aimed at enhancing the agriculture “industry” could be instituted (for example, providing all‐season access and feeder roads and establishing markets for products, to name a few).

This paper aims at examining and mapping the rainfall variability. This study confirmed that GIS kriging interpolation technique is practically helpful in assessing variation in climate parameters. There are a number of methods available for rainfall variability assessment (Tveito and Schöner, 2002), but their application is restricted by theoretical assumptions that must be fulfilled as far as possible. However, well‐constructed GIS maps allow presentation at a wide range of spatial scales and virtually unlimited regimens for data processing. At the same time, the construction of digital maps is now possible because of progress in computer science, data availability, exchange and access, and worldwide communication networks. GIS are unique because of their emphasis on providing users with a representation of objects in a cartographically accurate spatial system and on supporting analysis and decision makings (Burrough and Rachael, 1998).

Generally, the main findings in this paper are related to its methodology. It is easy to calculate and map the mean monthly or seasonal rainfall for specified periods and the deviations from the mean value. It is hard to imagine contemporary rainfall monitoring without GIS applications. The paper concluded that geospatial techniques are powerful tools that should be explored further for realistic assessment of the effects of climate variability on farming activities.

This paper is part of a research project funded and supported by Council for the Development of Social Science Research in Africa (CODESRIA). The author would like to thank the Space Application and Environmental Science Laboratory coordinated by Professor A.T. Salami, of Institute of Ecology, Obafemi Awolowo University, Ile‐Ife, Nigeria, for the scientific contribution to training activities and fellowship grant support.

Ayansina Ayanlade is a Researcher and Lecturer in the Department of Geography, Obafemi Awolowo University, Ile‐Ife, Nigeria. He is a professional in GIS, remote sensing, web cartography and has engaged in geospatial analysis of climate variability impacts research. Ayansina Ayanlade can be contacted at: sinaayanlade@yahoo.co.uk