Improved weighting methods, deterministic and stochastic data-driven models for estimation of missing precipitation records

Abstract

Distance-weighted and data-driven methods are extensively used for estimation of missing rainfall data. Inverse distance weighting method (IDWM) is one of the most frequently used methods for estimating missing rainfall values at a gage based on values recorded at all other available recording gages. In spite of the method's wide success and acceptability, it suffers from major conceptual limitations. Conceptual improvements are incorporated in the IDWM method that led to several modified distance-based methods. A data-driven model that uses artificial neural network concepts and a stochastic interpolation technique, kriging, are also developed and tested in the current study. These methods are tested for estimation of missing precipitation data. Historical precipitation data from 20 rain-gauging stations in the state of Kentucky, USA, are used to test the improvised methods and derive conclusions about the efficacy of incorporated improvements. Results suggest that the conceptual revisions can improve estimation of missing precipitation records by defining better weighting parameters and surrogate measures for distances that are used in the IDWM.