| Measuring drought and drought impacts in Red Sea Province |
|2. Measuring drought and food insecurity in Red Sea province: in 1987 and 1988: a technique for Pthe rapid assessment of large areas. Roy Cole|
Although expanding the complexity of the work goes counter to our intention to explore alternative, low-technology methods of areal assessment, this type of research would benefit from the high technology approach. The use of the Normalised Difference Vegetation Index (NDVI) maps rather than subjective assessments of rainfall and vegetation quality would have gone far to reduce error in classification. The NDVI is prepared on a monthly basis for the governments of most drought-prone African countries by the United States Geological Survey, National Mapping Division at the EROS Data Center in the United States. It is funded by the Africa Bureau of the United States Agency for International Development (USAID). NDVI images are produced from the United States National Oceanographic and Atmospheric Administration's Advanced Very High Resolution Radiometer (AVHRR) satellite data. The scale of the imagery is 1:2500000 and the resolution is 1 km2. Reflectance values of vegetation are on a greenness scale from -1.0 to 1.0, the lower limit being equivalent to bare rock and the upper to green fields of wheat.
The study would have benefited from the use of more variables, for example, rainfall in millimetres, cropped and harvested areas, human and livestock populations, number of migrant workers, wage rates and destinations, security, number of refugees, malnutrition rates, and locust damage to crops. These types of data together would go well in a principal components or factor analysis. Principal components and factor analysis are used to bring together variables and different types of data and create new variables that express the relationships between the original variables. These new variables can be mapped in meaningful ways. There are two interrelated reasons to use principal components or factor analysis in a study such as the present one4:
1. To identify groups of intercorrelated variables.
2. To create new variables through combination of old variables.
Intercorrelated variables are independent variables that measure the same thing in different ways. In combining the variables we define the "thing" we are measuring. Sometimes this is called "fishing around in the data." One variable, for example, human population, may not explain the distribution of livestock in Red Sea Province at a given time during the year. The aggregation of more than one variable will do so. For example, what is the relationship between the variables, rainfall, tribal affiliation, human population, livestock populations and the Braytek basin? It is a good guess that these variables may explain it. The factor analysis combines such variables and creates factors instead. Out of 25 variables, for example, the factors produced by factor analysis would be about 4. Each factor expresses a characteristic or a concept. For example, a factor composed of the variables, human population, livestock population, rainfall, and location possesses the characteristics of rainfall, rural, areas of good pasture production, people, and location. It may be called "areas of good grazing and human and livestock concentration". In Red Sea Province, where this variable is found to be high one can expect all that the variable measures to exist there. Where it is low, the opposite is expected. Other groups of variables may be so grouped. This reduces many of the headaches of determining significant variables and also creates important variables from variable that alone would not be significant.
The use of a Geographic Information System would also be a great help for similar studies that use more variables. A Geographic Information System (GIS) is a computer programme that stores, analyzes and makes maps of data such as those collected for the present study. A GIS has the power to manipulate masses of data, do much of the statistical work done by expensive statistics programmes (which, these days, may include the creation of factors from groups of variable), and automate the process from data to maps (see Clarke (1986) and Lillesand and Kiefer (1987) for further information). The investment necessary for the high technology approach would be more appropriate for large and long term programmes with generous budgets and secure financing. A prograrnme such as this is more appropriately conducted on the national or international level rather than at the provincial level. A system such as or similar to that described in the present paper would go a long way in providing useful data for drought impacts and food insecurity assessments on the regional or national level. The most important consideration in setting up such a system is to keep the unit of analysis, the cell or zone, as small as possible. If the unit of analysis is too big it becomes insensitive or masks the differences within the unit. Ideally, differences within a unit should be minimised. Good criteria in constructing a unit would be uniform economic activities, land use, land cover, ethnic, cultural or familial uniformity, and urbanisation or linkage to urban areas. Although the District Councils in the Sudan may be of an appropriate size and homogeneity for zoning in a meaningful way, there may be traditional systems of areal zoning that are more useful, for example, lineage or ethnic areas. In the case of the Sudan, it may be more convenient to use the District Councils because they are a unit of data collection on a variety of topics. For example, data are collected at the District Council headquarters on cultivated and harvested areas within the District Councils, rainfall and market prices for cereals and livestock. It is here that data for the Islamic social security system (from the zakaat) are collected. It should be noted that the District Council can be broken down into any number of units because of the way the data are collected. For example, data on cultivated and harvested areas are collected by khor. If a particular khor fits better into some other spatial grouping than the District Council, then it can be regrouped. Also available are data on rainfall (Meterological Department in Khartoum) and wadi flooding (National Water Corporation in Khartoum). Rainfall data are available on a daily, monthly, or annual basis. Available flood data, collected from permanent gauging stations around the country, concern the number of floods, the duration of each flood, the maximum discharge in cubic metres and the total flow per flood in cubic metres. In Red Sea Province there are seven such gauging stations The agricultural schemes throughout the Sudan collect their own data on rainfall and flooding as well. NDVI maps of vegetation on a monthly basis are available, as mentioned above, but in deeply incised areas of the Red Sea Hills their usefulness is limited because the large unit of analysis (1 km2) mixes the reflectance values for the khors with the mountains. Landsat Thematic Mapper data would be more appropriate but are more expensive and take longer to process than the NDVI data.