Developing a methodology for the inverse estimation of root architectural parameters from field based sampling schemes

Developing a methodology for the inverse estimation of root architecturalparameters from field based sampling schemesShehan Morandage (1), Andrea Schnepf (1), Jan Vanderborght (1), Mathieu Javaux (1,2), Daniel Leitner (3), EricLaloy (4), and Harry Vereecken (1)(1) Forschungszentrum Jülich, IBG-3, jülich, Germany (t.morandage@fz-juelich.de), (2) Earth and Life Institute, Universitécatholique de Louvain, Louvain,Belgium (m.javaux@fz-juelich.de), (3) Department of Forest and Soil Sciences, University ofNatural Resources and Applied Life Sciences, Vienna, Austria (daniel.leitner@univie.ac.at), (4) Institute for Environment,Health and Safety, Belgian Nuclear Research Centre, Belgium (eric.laloy@sckcen.be)Root traits are increasingly important in breading of new crop varieties. E.g., longer and fewer lateral roots aresuggested to improve drought resistance of wheat. Thus, detailed root architectural parameters are important.However, classical field sampling of roots only provides more aggregated information such as root length density(coring), root counts per area (trenches) or root arrival curves at certain depths (rhizotubes). We investigate thepossibility of obtaining the information about root system architecture of plants using field based classical rootsampling schemes, based on sensitivity analysis and inverse parameter estimation.This methodology was developed based on a virtual experiment where a root architectural model was usedto simulate root system development in a field, parameterized for winter wheat. This information provided theground truth which is normally unknown in a real field experiment. The three sampling schemes coring, trenching,and rhizotubes where virtually applied to and aggregated information computed. Morris OAT global sensitivityanalysis method was then performed to determine the most sensitive parameters of root architecture model forthe three different sampling methods. The estimated means and the standard deviation of elementary effects of atotal number of 37 parameters were evaluated. Upper and lower bounds of the parameters were obtained basedon literature and published data of winter wheat root architectural parameters. Root length density profiles ofcoring, arrival curve characteristics observed in rhizotubes, and root counts in grids of trench profile methodwere evaluated statistically to investigate the influence of each parameter using five different error functions.Number of branches, insertion angle inter-nodal distance, and elongation rates are the most sensitive parametersand the parameter sensitivity varies slightly with the depth. Most parameters and their interaction with the otherparameters show highly nonlinear effect to the model output.The most sensitive parameters will be subject to inverse estimation from the virtual field sampling data us-ing DREAMzs algorithm. The estimated parameters can then be compared with the ground truth in order todetermine the suitability of the sampling schemes to identify specific traits or parameters of the root growth mode