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

Precipitation data from rain gauges is fundamental across many lines of enquiry within hydrology and environmental science. A widespread problem with data sourced from remote sensors is that of missing data and gaps in established datasets. The accurate recovery of missing data is critical to maximise information gain from data sources, with requirements for environmental data at high temporal resolution ever increasing. Precipitation data collected at sub-hourly resolution represents specific challenges for data recovery by being largely stochastic in nature and highly unbalanced in 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.