Evaluation of Improvised Spatial Interpolation Methods for Infilling Missing Precipitation Records

Historical rainfall records are of extreme importance to most hydrologic analyses, including water resources evaluation, impact of land use changes, urban runoff volumes and durations, hydro periods, flood analyses, climate variability, among many others. However, daily precipitation records are rarely complete and suffer from occasional and extended periods of missing values. This study involves evaluation of deterministic and stochastic interpolations methods to fill the gaps in daily precipitation records. Evaluated methods include four improvised methods as well as three benchmark methods, with implementation variations for all methods. Benchmark methods are selected to provide a basis against which the evaluated methods performance can be quantified and compared. Evaluated methods include variations of Inverse Distance Weighting Method (IDWM), Correlation Coefficient Weighting Method (CCWM), Linear Weight Optimization Method (LWOM) and Artificial Neural Networks (ANNs). Benchmark methods are variations of the Single Best Estimate (SBE) method and spatial and climatological mean estimators (SME and CME). In addition, two classification methods were evaluated in combination with the estimation methods in an attempt to separate zero and non-zero precipitation estimates. The evaluated classifiers are Support Vector Machine (SVM) and variations of a Single Best Classifier (SBC). The methods are tested for application at 53 NOAA rain gauges that cover areas of interest to the Southwest and South Florida Water Management District regions in South Florida.