Appropriate conventional methods for estimating missing precipitation values in Sri Lanka

Abstract

This study focused on finding the most appropriate conventional method for estimating missing precipitations values in different climate zones in Sri Lanka. The main location and surrounding locations were selected from each zone considering their data availability. Monthly precipitation data for each selected locations were collected from 1980 to 2020. Three temporal methods; Holt Winter’s Exponential Smoothing, Seasonal Auto Regressive Integrated Moving Average and, Multiple Aggregation Prediction Algorithm and four spatial methods; namely, Arithmetic Mean Method (AMM), Normal Ratio Method (NR), Inverse Distance Weightage Method (IDW) and Aerial Precipitation Ratio (APR) method were used for estimating missing values. Root Mean Square Error, Mean Absolute Error, and Mean Absolute Percentage Error values were used for selecting the most appropriate method for estimating missing values for each location. It was difficult to name one method as the most appropriate method for all climate zones. Spatial methods have performed better than temporal methods in all seven climate zones. The missing values in the Low Country Wet Zone and Up Country Wet Zone could be estimated using AMM and NRM methods. APR method is the most suitable for estimating missing values in the Mid Country Wet Zone and Low Country Intermediate Zone. IDM method performed well in Mid Country Intermediate Zone, Up Country Intermediate Zones and Low Country Dry Zone. However, further improvement of both temporal and spatial methods for estimating missing values is necessary as none of these methods are capable of giving estimates with high accuracy.