Calibration of DART 3D model with UAV and Sentinel-2 for studying the radiative budget of conventional and agro-ecological maize fields

Quantifying the radiative budget (RB) of maize, the world's predominant cereal, is crucial to managing its growth in a context of global warming. Its microclimate including temperature, water, and energy flux depend on crop management practices. This paper compares the influence of agroeco-logical practices (AE) and conventional agriculture (CA) on the RB of maize using three-dimensional (3D) radiative transfer modelling and remote sensing (RS) observations. We studied two neighbour maize fields in south-west France, respectively with AE and CA practices. Time series of photo-synthetically active radiation absorbed (APAR) by plants (APARplant) and soil (APARsoil) were simulated with the DART radiative transfer model. First, realistic 3D models of the CA and AE fields were created and validated with a new method based on DART and Sentinel 2 and UAV reflectance and vegetation indices acquired in July 2019. At that date, around midday, APARplant was larger by 21.5 W/m2 in the AE field and APARsoil was larger by 20.1 W/m2 in the CA field. We explained these differences by differences in geometric and optical parameters (OP) between the two fields. OPsoil causes 45% of the differences in APARplant and APARsoil. The shape of plant causes 14% of the difference in APARplant and 17% in APARsoil. OPplant causes 14% of the difference in APARplant and 1% in APARsoil. Field geometry (row orientation, inter-row and plant distance) causes 7% of the difference in APARplant and 2% in APARsoil. It stresses the great role of OPsoil associated to AE practices. Because it integrates the non-linear ef-fects of all above parameters, the LAI (Leaf Area Index) accounts for 40% difference in APARplant and APARsoil. The sensitivity of APAR to LAI is twice as low in the AE field than in the CA field. We showed that APAR differences between the AE and CA fields can be even greater with other maize row orientations. We also highlighted the role of the field model on simulated APAR by comparing the use of APAR simulated by a 3D field model and its corresponding 1D turbid field: 1D turbid models resulted in a 16% increase in APARplant and in similar APARsoil, compared to the 3D CA and AE models.