118 Forage mass estimation of sudangrass-based pasture using vegetation indices.

Le résumé fourni par la source

Abstract Remote sensing (e.g., satellites), can detect the electromagnetic radiation reflected by plants and translate it into numerical values. The values from different spectral bands can be combined mathematically into vegetation indices (VI) that have been used to assess plant health, activity, and biomass. The objective of this study was to assess ability of five different VI to predict forage mass of sudangrass-based pastures when rotationally stocked. The five VI used were: Normalized Difference Vegetation Index (NDVI), Spectral Feature Depth Vegetation Index (SFDVI), Green Leaf Index (GLI), Red Edge Chlorophyll Index (RECI), and Normalized Difference Red Edge Index (NDRE). This project was conducted over two summers with a total of 63 estimates of forage mass being taken prior to cattle entry into a paddock (1.6 ha) with 36 being collected in year 1 and 27 in year 2. The paddocks contained either a monoculture of sudangrass (sorghum × drummondii) or a mixture of sudangrass and sunnhemp (Crotalaria juncea), with 33 estimates taken from the monoculture and 30 from the mix. Sunnhemp contributed only 10% of the forage mass in the mix. The forage mass within each paddock was estimated by clipping four randomly selected areas (0.49 m²) at ground level. The standing height of the forage was recorded at 20 random points within each paddock. The forage mass ranged from 1,382 to 9,648 kg DM/ha, with a mean of 3,418 kg DM/ha. PlanetScope satellite images corresponding to the forage sampling days were obtained via Planet Explorer and VIs were calculated using QGIS. Linear regression was performed using PROC REG in SAS. When standing height was regressed with forage mass, adjusted R² was 0.63 and root mean square error (RMSE) was 718 kg DM/ha, suggesting standing height was a good predictor of mass. When each of the five vegetation indices was regressed with mass separately, adjusted R² was negative and the RMSE was large (≥ 1,441 kg DM/ha), suggesting that these VIs alone were not good predictors of forage mass. A multiple regression analysis was performed between forage mass and all five VIs using the random forest regression algorithm. However, once again the fit was poor with an R² = -0.10 and RMSE = 1,092 kg DM/ha. In an attempt to refine this multiple regression, standing height was included as another input variable but did not improve the prediction of forage mass (R² = 0.25 and RMSE = 902 kg DM/ha) over standing height alone. These results suggest that, for sudangrass-based pastures, these VIs are not good predictors of forage mass. This is possibly due to the plant morphology, as these plants had a low leaf-to-stem ratio as the plant matured later in the grazing season.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.