Optimal functional forms for estimation of missing precipitation data

A fixed functional set genetic algorithm method (FFSGAM) is proposed and is investigated in the current study to obtain optimal functional forms for estimating missing precipitation data. The FFSGAM provides functional forms with optimal combination of parameters of surrogate and actual measures of strength of correlation among observations for estimating missing data. The method uses genetic algorithms and a nonlinear optimization formulation to obtain optimal functional forms and coefficients, respectively. Historical daily precipitation data available from 15 rain gaging stations from the state of Kentucky, USA, are used to test the functional forms and derive conclusions about the efficacy of the proposed method for estimating missing precipitation data. The tests of FFSGAM at two rainfall gaging stations in Kentucky, using multiple error and performance indices, indicate that better estimates of precipitation can be obtained compared to those from a traditional inverse distance weighting technique. Also, results from the use of the method confirm its robustness when only six rain gaging stations out of 14 were used for estimating missing data.