THESIS DEFENSE of Vu Hoang Ha PHAM - December 1, 2025

This thesis investigates the spatio-temporal variability of agricultural fields by combining two sources of imaging data: fixed sensors providing temporal series, and mobile sensors delivering spatial maps. The objective is to assess the relevance of this data fusion for reconstructing field dynamics—applied here to vineyards using leaf area as the variable of interest—as well as the sampling effort required for an effective reconstruction. Two approaches were implemented: a controlled simulation integrating the main sources of variability, and an experiment conducted on an instrumented vineyard. Three reconstruction methods are compared: spatio-temporal inverse distance weighting (STIDW), spatio-temporal kriging (STK), and generalized additive models (GAM). The results show that STK and GAM successfully capture spatio-temporal variability, and that reconstruction quality depends on the number of sensors and acquisitions. An experimental trade-off emerges, relying on 4 to 9 fixed sensors per hectare and a few mobile acquisitions early in the season.