A data-driven model of the state of the African biosphere Martin Jung, Eric Thomas
Meteorology Land use Biosphere- atmosphere exchange (FLUXNET) Biosphere- atmosphere exchange (FLUXNET) Vegetation state from satellites Empirical ‘upscaling’ of FLUXNET Pinty et al 2011 Explanatory (X) variablesTarget (Y) variable
Meteorology (7 x 4) Land use (8) Soil (10) Remotely sensed fAPAR Remotely sensed fAPAR Mean annual Mean seasonal cycle Anomalies Raw Random forests Empirical modelling of phenology Explanatory (X) variablesTarget (Y) variable
Approach Training with all 48 variables Time period: Prediction of fAPAR for Evaluation of predicted fAPAR Case Study Africa
Results
Forecasting of anomalies Results
Improving modelling of anomalies Meteorology (7 x x 24 ) Land use (8) Soil (10) Remotely sensed fAPAR Remotely sensed fAPAR Mean annual Mean seasonal cycle Anomalies Raw Random forests Lag Cumulativ e Lag Lag Cumulativ e Lag Fire fAPAR Mean season al cycle Variance of anomalies Variable selection based on Genetic Algorithm Back to methods …
Forecasting of anomalies improved oldnew New Results
Potential applications -In ‘early warning systems’ by using seasonal weather forecasts -In empirical ‘upscaling’ of FLUXNET data beyond satellite era -In Land Surface Models Development of a fully data-driven phenology model