@misc{10481/72317,
year = {2022},
month = {3},
url = {http://hdl.handle.net/10481/72317},
abstract = {This study proposes a new methodology to estimate the Atmospheric Boundary Layer Height (ABLH), discriminating between Convective Boundary Layer and Stable Boundary Layer heights, based on the machine learning algorithm known as Gradient Boosting Regression Tree. The algorithm proposed here uses a first estimation of the ABLH derived applying the gradient method to a ceilometer signal and several meteorological variables to obtain ABLH values comparable to those derived from a microwave radiometer. A deep analysis of the model configuration and its inputs has been performed in order to avoid the model overfitting and ensure its applicability. The hourly and seasonal values and variability of the ABLH values obtained with the new algorithm have been analyzed and compared with the initial estimations obtained using only the ceilometer signal. Mean Relative Errors (MRE) between the ABLH estimated with the new algorithm and microwave radiometer show a daily pattern with their highest values during the night-time (stable situations) and their lowest values along the day-time (convective situations). This pattern has been observed for all the seasons with MRE ranging between −5% and 35%. This result notably improves those ABLH values derived by applying the gradient method to ceilometer data during convective situations and enables the Stable Boundary Layer height detection at night and early morning, instead of only Residual Layer top height. Finally, the model performance has been directly validated in three particular cases: clear-sky day, presence of low-clouds and dust outbreak event. In these three particular situations, ABLH values obtained with the new algorithm follow the pattern obtained with the microwave radiometer presenting very similar values, thus confirming the good model performance. In this way it is feasible by the combination of the proposed method with gradient method, to estimate Convective, Stable and Residual Boundary Layer height from ceilometer data and surface meteorological data in extended network that include ceilometer profiling.},
organization = {Spanish Ministry of Economy and Competitiveness through projects CGL2015-73250-JIN, CGL2016-81092-R, CGL2017-83538-C3-1-R, CGL2017-90884-REDT and PID2020-120015RB-I00},
organization = {COST Action TOPROF (ES1303), supported by COST (European Cooperation in Science and Technology)},
publisher = {Elsevier},
keywords = {Atmospheric Boundary Layer height},
keywords = {Ceilometer},
keywords = {Gradient boosting regression tree},
title = {Estimating the urban atmospheric boundary layer height from remote sensing applying machine learning techniques},
doi = {https://doi.org/10.1016/j.atmosres.2021.105962},
author = {Arruda-Moreira, G. and Sánchez Hernández, Guadalupe and Guerrero Rascado, Juan Luis and Cazorla Cabrera, Alberto and Alados Arboledas, Lucas},
}