Water, Energy, and Carbon with Artificial Neural Networks
(WECANN): A statistically-based estimate of global surface turbulent
fluxes using solar-induced fluorescence
Abstract. A new global estimate of surface turbulent fluxes, including latent heat flux (LE), sensible heat flux (H), and gross primary production (GPP) is developed using remotely sensed Solar-Induced Fluorescence (SIF) and other radiative and meteorological variables. The approach uses an artificial neural network (ANN) with a Bayesian perspective to learn from the training datasets: a target input dataset is generated using three independent data sources and a triple collocation (TC) algorithm to define a prior distribution. The new retrieval, named Water, Energy, and Carbon with Artificial Neural Networks (WECANN), provides surface turbulent fluxes from 2007 to 2015 at 1° × 1° spatial resolution and on monthly time resolution. The quality of ANN training is assessed using the target data, and the WECANN retrievals are validated using FLUXNET tower measurements across various climates and conditions. WECANN performs well in most cases and is strongly constrained by SIF information. The impact of SIF on WECANN retrievals is evaluated by removing it from the input dataset of the ANN, and it shows that SIF has significant influence, especially in regions of high vegetation cover and in humid conditions. When compared to in situ eddy covariance observations, WECANN typically outperforms other estimates, particularly for sensible and latent heat fluxes.