[Post-doc completed] Leire Sandonis-Pozo: A hybrid approach to combining biophysical modeling and remote-sensing

A hybrid approach to combining biophysical modeling and remote-sensing derivatives model 3D canopy architecture in vineyards for differential management

Measuring the size, shape and density of vineyard canopies is not easy. The canopy will continually evolve over the season and the 3D characteristics of the canopy will strongly influence the need for and the efficacy of crop protection actions during the season. Proximal sensors can provide high-resolution information on the 3D canopy structure, but are limited in the spatial and temporal resolution of deployment. Satellite imagery provides high-resolution spatio-temporal information on vineyard vigour, but the image information is only partly influenced by canopy shape. Therefore, neither sensing system, in its native form, is capable of providing relevant vineyard information to support in-season, differential crop protection strategies. This project will use derivatives from remotely sensed imagery, obtained via time-series analysis of vegetation indices and from inverse radiative transfer models, to generate models that can predict the 3D characteristics of vineyards over a large area. The calibration and validation data for these models will be derived from high-resolution LiDAR data at selected points. The processing of these LiDAR data will be based on new algortihms that provide 3D information on the vineyard canopies. The modelling of the 3D canopy characteristics with the remote-sensing derivatives will be performed using a mixture of linear and non-linear machine-learning methods. Once a stable model has been found, the predicted, large-scale 3D canopy information will be substituted into exisiting spray deposition models to evaluate if the 3D canopy predictions are of sufficient quality to be used for management.
 The project will be built of existing data sets (Sentinel 2 imagery and LiDAR campaigns by UMR ITAP) and existing programmed algorithms for data processing (time-series, LiDAR and radiative transfer models from both UMR ITAP and TETIS). It seeks to connect several previous #DigitAG projects to generate useful, operational vineyard information.

Contact: leire.sandonis(at)udl.cat

Réseaux : Linkedin