Eddy covariance flux data for supporting local C02 emissions reduction strategies: integration of direct observations, remote sensing and a machine learning approach