Estimating the urban atmospheric boundary layer height from remote sensing applying machine learning techniques

This study proposes a new methodology to estimate the Atmospheric Boundary Layer Height (ABLH), discriminating between Convective and Stable heights, based on machine learning algorithm known as Gradient Boosting Regression Tree. The proposed here uses first estimation of ABLH derived applying gradient method ceilometer signal several meteorological variables obtain values comparable those from microwave radiometer. A deep analysis model configuration its inputs has been performed in order avoid overfitting ensure applicability. hourly seasonal variability obtained with have analyzed compared initial estimations using only signal. Mean Relative Errors (MRE) estimated radiometer show daily pattern their highest during night-time (stable situations) lowest along day-time (convective situations). observed for all seasons MRE ranging ?5% 35%. result notably improves by data convective situations enables height detection at night early morning, instead Residual top height. Finally, performance directly validated three particular cases: clear-sky day, presence low-clouds dust outbreak event. In these situations, follow presenting very similar values, thus confirming good performance. this way it is feasible combination method, Convective, surface extended network that include profiling.