Table 1.
Pearl millet varieties used in this study.
Fig 1.
Root intersections density (RID) counting method used for root length density (RLD) modeling from RID.
(A) Experimental design and trench profile for root sampling (at 30 cm from the plant in this example), (B) and (C) sampling device with sides oriented according to the soil surface and plant row (H: horizontal, L: longitudinal, T: transversal), (D) and (E) root sampling process, (F) Root impacts counting on all three sides of soil cubes extracted from the trench profile, and (G) grid (5x 5 cm mesh) on a soil profile for soil-roots intersections counting (RI).
Fig 2.
Characteristics of the varieties used for model calibration.
(A) Root length density, (B) root biomass density and (C) shoot biomass. Data are mean +/- standard deviation. Significant differences (Tukey’s HSD) are indicated by different letters. For root traits, the mean of the two observation dates (60 DAS and 80 DAS) was considered.
Fig 3.
Fine and thick root growth orientation.
Relationship between the number of measured root impacts on a vertical face and measured root length density.
Fig 4.
Elaboration of geometric models (all, thick and fine roots).
Relationship between soil depth measurements (meters) and the main direction of root growth in relation to a vertical plane (Pv index).
Table 2.
Models validation analyses.
Fig 5.
Test of the relationship between measured and calculated RLD for the four proposed models.
(A) Geometrical models for fine roots (diameter<1 mm) and (B), thick roots (diameter >1 mm). (C) Empirical model with all varieties bulked and (D) geometrical model with all roots For all models, a prediction interval of 95% is indicated by the green lines. It gives an estimate of the interval in which a future observation will fall with a certain confidence level (here 95%), given previous observations used to build the model.
Fig 6.
Agromorphological characteristics of SL28 and LCICMB1.
(A) Shoot biomass (g.m-2), and (B) total grain weight (g.m-2) measured at the end of cycle for WW and DS conditions.
Fig 7.
Impact of water deficit on root length density distribution according to depth in SL28 and LCICMB1.
SL28 at 43 DAS (A), and 71 DAS (C), LCICMB1 at 44 DAS (B), and 72 DAS (D). Mean of RLD of four plants per variety was considered for each variety.
Fig 8.
Impact of water deficit on mean root distribution for SL28 and LCICMB1 in the soil profile.
Data mapped on a 0.05 × 0.05 m grid like in the field and expressed in root length density (RLD) in WW and DS conditions for SL28 in (A) and the inbred line LCICMB1 in (B).
Fig 9.
Impact of water deficit on total root length of SL28 and LCICMB1.
(A) Total root length (m.m-3) measured at 72 DAS at the end of water stress treatment, and (B) ratio between total root length (m.m-3) and shoot biomass.