Evaluation of infilling methods for time series of daily precipitation and temperature: The case of the Ecuadorian Andes

Abstract

Continuous time series of precipitation and temperature considerably facilitate and improve the calibration and validation of climate and hydrologic models, used inter alia for the planning and management of earth’s water resources and for the prognosis of the possible effects of climate change on the rainfall-runoff regime of basins. The goodness-of-fit of models is among other factors dependent from the completeness of the time series data. Particular in developing countries gaps in time series data are very common. Since gaps can severely compromise data utility this research with application to the Andean Paute river basin examines the performance of 17 deterministic infill methods for completing time series of daily precipitation and mean temperature. Although sophisticated approaches for infilling gaps, such as stochastic or artificial intelligence methods exist, preference in this study was given to deterministic approaches for their robustness, easiness of implementation and computational efficiency. Results reveal that for the infilling of daily precipitation time series the weighted multiple linear regression method outperforms due to considering the ratio of the Pearson correlation coefficient to the distance, giving more weight to both, highly correlated and nearby stations. For mean temperature, the climatological mean of the day was clearly the best method, most likely due to the scarcity of weather stations measuring temperature, and because the few available stations are located at different elevations in the landscape, suggesting the need to address in future studies the impact of elevation on the interpolation.