Evaluating Inverse Distance Weighting and Correlation Coefficient Weighting Infilling Methods on Daily Rainfall Time Series
Year: 2021
Type: article
Abstract: Encountering missing daily rainfall records is inevitable and filling the gap is a challenging issue. Spatial interpolation is one of the most widely used methods to estimate missing daily rainfall. The method is easy to apply, less time consuming, and requires inexpensive computation than more complex methods. This study attempted to evaluate the inverse distance weighting (IDW) and correlation coefficient weighting (CCW) methods and compare each method’s performance over the Upper Ping River Basin. Daily rainfall data from 92 stations over about 65 years (1953 – 2017) were obtained. After screening, 44 stations were used in this study. Before implementing infilling methods, cluster analysis using K-means was applied to group the station into three sub-regions. Three and four neighbor stations (source stations, SSs) were tested to find the optimal number of SSs. Six target stations from three sub-areas were chosen to test the infilling method with different percentages of missing values to represent various numbers of missing data. The study results revealed that CCW provided better performance than the IDW method. The optimal number of source stations to estimate missing data was four stations assessed by evaluating mean values, R correlation and similarity index. Moreover, CCW also yielded less error of mean absolute error (MAE) and root means square error (RMSE) compared to IDW. Varying the percentage missing values between 5%, 10%, 20%, 30%, 40%, and 50% revealed that each infilling method was not sensitive to the percentage of missing data.
Source: Creative Science
Institution: Burapha University
Language: English
Subfield: Atmospheric Science
Field: Earth and Planetary Sciences
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7216675770