Fig 1.
Map of the study area with 30 pasture plots, five bare soil plots and six control plots.
Table 1.
Vegetation indices tested for biomass estimation using 12 biomass values from the control unit, range of values observed (minimum and maximum values) and selection criteria: Coefficient of variation (CV) and regression quality (coefficient of determination (R2) of the linear regression between predicted and observed values for fresh (F) and dry (D) biomass).
Fig 2.
Location of samples used for GNDVI-based biomass model and vegetation cover classification.
Table 2.
Number and characteristics of biomass quadrats (0.5 × 0.5 m) collected in the field during the growing season and used for the GNDVI-based biomass estimation model.
Fig 3.
Linear regressions between volumetric values and fresh (a) and dry (b) biomass (n = 12).
Fig 4.
Non-linear regressions between GNDVI values and fresh (a) and dry (b) biomass (n = 49).
Fig 5.
Concordance analysis between observed and predicted values for fresh (a) and dry (b) biomass of the validation dataset (n = 50). Continuous line = 1:1 line, dashed line = linear regression.
Fig 6.
Values for average eGNDVI from polygons selected in bare soil and pasture units during the entire growing season and for different duration of forage growth (n = 248).
Fig 7.
Visual comparison of three biomass or vegetation cover estimation methods for the imagery acquired on August 29th in the control unit: (a) the RGB imagery including the forage growth duration per plot (weeks), (b) the volumetric-based biomass model, (c) the GNDVI-based biomass model, and (d) the GNDVI-based vegetation cover classification.