Rainfall and cotton yields in the Sudan Gezira

Abstract

Agricultural meteorology makes slow progress because it is rarely possible to study the correlation of weather and yield for crops grown in normal agricultural rotations over long periods. Both in commercial agriculture and at experiment stations, rotations and cultural methods change sufficiently frequently to destroy the homogeneity of the data, even when records are kept for specific fields or farms. In some of the newer agricultural areas, and especially in those under irrigation, the cropping is simpler, and the need for records and field experiments is more keenly appreciated. Research programmes and practical measures must be determined from experience over only a few decades; but erroneous conclusions may easily be drawn from trends over such short periods, memories of individual years, or simple correlations of yields and single weather variates. In the present paper, modern statistical methods of analysing small samples of data are used to examine the relationships between rainfall and cotton yields under irrigation in the Sudan Gezira for periods up to 23 years. The mean yields from this large and semi-desert area fluctuate violently from year to year in such a way as to suggest the dominance of some climatic factor. It will be shown that the cotton yields are highly correlated with rainfall, and further, that seasonal fluctuations in rainfall, including certain cyclic changes, are sufficient to account for most of the decline in yield on some of the oldest areas. This conclusion has considerable practical importance, because a recent sequence of low yields led some critics (Balls, 1935, Vageler and Alten, 1932) of the scheme to conclude that irrigation without drainage had already produced serious soil deterioration. The significance of some of the rainfall effects established may direct attention to new methods of studying problems of soil fertility and the control of pests.