CO2 Gap filling

Two methods were used for reconstruction: 1) direct recovery based on available episodic data; 2) reconstruction using multidimensional linear regression on meteorological data. Both methods are based on machine learning.

The first method uses CO2 flux and datetime data. The second method, in addition to CO2 flux and datetime, uses parameters such as air temperature, vapour pressure deficit [VPD], surface latent heat flux [LE], surface sensitive heat flux [H], relative humidity [RH], dynamic wind speed [u*], short-wave (solar) radiation entering the the surface is [SW_IN_1_1_1]. To obtain the missing meteorological parameters, data from the ERA5 reanalysis was used, from the dataset "ERA5 hourly data on single levels from 1940 to present".