Assessing Machine Learning Models for Gap Filling Daily Rainfall Series in a Semiarid Region of Spain

The presence of missing data in hydrometeorological datasets is a common problem, usually due to sensor malfunction, deficiencies records storage and transmission, or other recovery procedures issues. These values are the primary source problems when analyzing modeling their spatial temporal variability. Thus, accurate gap-filling techniques for rainfall time series necessary have complete datasets, which crucial studying climate change evolution. In this work, several machine learning models been assessed gap-fill data, using different approaches locations semiarid region Andalusia (Southern Spain). Based on obtained results, use neighbor located within 50 km radius, highly outperformed rest approaches, with RMSE (root mean squared error) up 1.246 mm/day, MBE (mean bias ?0.001 R2 0.898. Besides, inland area results coastal most locations, arising efficiency effects based distance sea (up an improvement 63.89% terms RMSE). Finally, (ML) (especially MLP (multilayer perceptron)) notably simple linear regression estimations sites, whereas improvements were not such significant.