STATISTICAL PREDICTION OF GULLY EROSION DEVELOPMENT ON THE COASTAL PLAIN SANDS OF THE SOUTH EASTERN NIGERIA

Abstract

A statistical model for predicting gully initiation was developed using variables from 20 randomly selected sites. The random samples gave 10 gully sites and 10 non-gully sites. In all, 12 variables were identified but using students t-tests approach, only four variables contributed to gully development. These four variables include Maximum slope, Maximum slope length, Microrelief amplitude and percentage coarse sand. The four variables were combined through factor analysis and statistical manipulations to form Linear Discriminant Function (LDF). Three functions were obtained by combining the variables in three different ways. An application of the three functions to the field situation identified function 1,g1 as a very comfortable prediction. When Yl was used to classify the various sites using the variables obtained from the field, a 25% wrong classification was obtained. This value was quite low when compared with the other two functions whose wrong classification ranged from 35% and above. It was observed from the study that when Yl is less than 30, it indicated little or no gully erosion threat.