Minimizing acute to severe risks from floods calls for efficient risk management strategies that include preparedness and active response. Predicting flash floods in an area is one form of determining the amount of associated risk thus being a basis for preparedness. Since the focus of my study is to determine whether the amount of rainfall in September in the coastal cities can be used to predict flash floods in another city, kriging interpolation method can be employed to estimate rainfall values in another city. More particularly, the study can apply ordinary interpolation kriging that provides probabilities under which predictions of a variable occurring lie. The interpolation method uses data from Geographic Information System.
The R statistical software will be employed to perform the kriging statistical analysis technique. Before conducting ordinary kriging, the data must be devoid of any skewness to reduce biasness. I will first create a semiva-riogram to visualize the variance in the estimates with distance between the sampled cities. Since ordinary kriging takes the assumption of spatial autocorrelation, the plot shows the range and distribution of data points across the cities.
Creation of several kriging models having varying semi-variogam models and numbers of lags is useful to compare the models’ accuracy. The root mean square error and the mean error are evaluated to give the uncertainty measure of the location that is not sampled. The best model should have a standardized mean that is close to zero. Similarly, it should have the least root-mean-square prediction error and the standard root-mean-square prediction error close to 1. The best fit semi-variogram parameters are employed on the datasets of the coastal cities that are sampled.The best model provides an interpolated layer of the area that is vulnerable to flash floods. The surface that is generated through interpolation is restricted within a specific buffer distance.