Bias correction of ERA5-Land air temperature in remote tropical regions using participatory monitoring data

Abstract

Many remote tropical mountainous regions lack precise, spatially distributed air temperature data. This limits both scientific research and applied solutions in fields such as hydrology, meteorology, health and agriculture. As regular monitoring networks do not provide sufficient spatial coverage in these regions, alternatives are required. To address this issue, this study proposes a novel approach using participatory monitoring (PM) data for point-to-point bias correction of ERA5-Land air temperature data. ERA5-Land data was trained on PM data from different remote mountainous regions in Ecuador (n = 67), Honduras (n = 23) and Tanzania (n = 275) using simple linear regression models. Validation was conducted using automatically measured air temperature, applying different metrics, including mean absolute error (MAE) and coefficient of determination (R2). Validation demonstrated an improvement across all stations, with the most significant reduction in MAE, from 5.49 °C to 1.76 °C in Ecuador. Concurrently, R2 increased across all stations up to 0.83. The decrease in deviation was also statistically significant for all stations. The study demonstrated that incorporating PM data into simple linear regression bias correction can reduce the regional bias of ERA5-Land air temperature data. This can be regarded as a pragmatic solution for remote regions, where the absence of weather stations is a common issue.