Fig 1.
Schematic representation of the integration of a UAV-modeling system.
Fig 2.
UAV-derived NDVI maps of grazed pasture of the KBS dairy farm (area of 15.6 ha) collected on four consecutive dates.
Maps exhibit the spatial variability of high-resolution measurements (6 cm pixels) of pasture NDVI (-1: blue to 1: red color) in two blocks of eight 1-ha paddocks (200 x 50m) used for rotational grazing with lactating dairy cows.
Table 1.
Characteristics of three methods of pasture monitoring used at the KBS dairy farm.
Technical data of Ruler, C-Dax and UAV based methods of pasture monitoring used during four consecutive weekly dates at the KBS dairy farm. Values in table represent reference data for a 1-ha paddock (200 x 50 m).
Fig 3.
Relationship among UAV-remotely sensed NDVI and pasture biomass (a) and leaf area index (b) in georeferenced plots of tall fescue- (black) and ryegrass-based (white) pasture plots (n = 72) in Exp. 1.
Fig 4.
Evaluation of the UAV method for estimation of biomass of tall fescue- (solid symbols) and ryegrass-dominated (open symbols) pasture for four sampling dates conducted during Exp. 2. (a) Graphical representation of goodness of fit of pasture biomass estimated by the UAV compared to mean reference values estimated by the two control methods, ruler and C-Dax. (b) Deviation of pasture biomass estimations by the UAV method compared to the ruler (■□) and C-Dax (▲Δ) methods.
Table 2.
Summary of statistics and comparison of the UAV-based method for estimation of herbage mass of tall fescue- and ryegrass-based pastures rotationally grazed by lactating dairy cows.
Comparisons were performed for each 1-ha paddock during the four pastures’ monitoring dates in Exp. 2.
Fig 5.
Predicted vs. observed leaf stage (a), leaf blade length (b) and digestibility of dry matter (c) of spring and summer regrowth of tall fescue- (solid symbols) and ryegrass-based (open symbols) pasture managed under three residual sward height treatments (low, ●○; medium, ■□; high, ▲Δ) in Exp.1.
Table 3.
Summary of statistics for testing of the MDP model for prediction of leaf morphogenetic traits and forage nutritive value of tall fescue- and ryegrass-based pasture.
Mean, standard deviation (SD), mean bias, root mean square error (RMSE), relative error (RE), coefficient of the regression (R2) and Pearson correlation coefficient (r) for leaf length (cm), leaf stage (leaves per tiller), neutral detergent fiber content (NDF, %), and digestibility of dry matter (DMD, %) and NDF (NDFD, %), both of summer and spring regrowth managed with different residual sward heights in Exp. 1.
Fig 6.
Example (Flight 4) of output map of estimated pasture biomass for a platform of 15.6 ha rotationally grazed with lactating cows in Exp. 2.
The map shows high-resolution (6 cm) spatial variability of pasture cover with colors denoting <500 kg DM ha-1 (red), 500–1700 kg DM ha-1 (yellow), 1700–2800 kg DM ha-1 (green) and > 2800 kg DM ha-1 (blue).
Fig 7.
Main outputs of the UAV-modeling system for Flight 4 (corresponding to Fig 6).
(a) Actual pasture cover and model estimation of digestibility (DMD, %) for each paddock in rotation. (b) Model estimations showing the predicted post-grazing (shaded bars) and both the predicted next pre-grazing biomass, digestibility (DMD, %) for each paddock in rotation. (c) Model estimations of leaf stage (leaves per tiller) and number of resting days needed to achieve the predefined pre-grazing target of 2700 kg DM ha-1 for all paddocks in rotation. The dotted line indicates the ideal resting days in relation to the length of rotation.
Fig 8.
Schematic representation of the leaf morphology and leaf stage of tillers simulated by MDP model.
Figures correspond to the average 1.3- and 4.0-leaf stage of plants recommended for tall fescue and ryegrass in order to achieve the predefined pre-grazing target of 2700 kg DM ha-1.