Imputation of missing sub-hourly precipitation data in a large sensor network: A machine learning approach

• Published in: Journal of hydrology (Amsterdam) Vol. 588; p. 125126
• Main Authors: Chivers, Benedict D., Wallbank, John, Cole, Steven J., Sebek, Ondrej, Stanley, Simon, Fry, Matthew, Leontidis, Georgios
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.09.2020
• Subjects: Data imputation; Environmental sensor networks; Gradient boosted trees; Machine learning; Precipitation; Soil moisture; Precipitation; Gradient boosted trees; Data imputation; Soil moisture; Machine learning; Environmental sensor networks
• ISSN: 0022-1694; 1879-2707
• DOI: 10.1016/j.jhydrol.2020.125126

•Missing rain data at sub-hourly resolution is recovered using a two-step analysis.•Data are first modelled as rain/no rain, and subsequently regressed as true values.•Models are informed by related hydro-meteorological and precipitation data.•Machine learning analyses perform better than surface fitting for data recovery.•Machine learning techniques allow for real-time data recovery at high resolution. Precipitation data collected at sub-hourly resolution represents specific challenges for missing data recovery by being largely stochastic in nature and highly unbalanced in the duration of rain vs non-rain. Here we present a two-step analysis utilising current machine learning techniques for imputing precipitation data sampled at 30-minute intervals by devolving the task into (a) the classification of rain or non-rain samples, and (b) regressing the absolute values of predicted rain samples. Investigating 37 weather stations in the UK, this machine learning process produces more accurate predictions for recovering precipitation data than an established surface fitting technique utilising neighbouring rain gauges. Increasing available features for the training of machine learning algorithms increases performance with the integration of weather data at the target site with externally sourced rain gauges providing the highest performance. This method informs machine learning models by utilising information in concurrently collected environmental data to make accurate predictions of missing rain data. Capturing complex non-linear relationships from weakly correlated variables is critical for data recovery at sub-hourly resolutions. Such pipelines for data recovery can be developed and deployed for highly automated and near instantaneous imputation of missing values in ongoing datasets at high temporal resolutions.