218 Comparison between vegetation indices to estimate biomass in cereal rye

Abstract

Listen

Translate article

Abstract Accurate forage mass estimation can lead to improved pasture management, but it requires a significant investment of time on the ground sampling. The use of satellite images has potential to reduce the time needed and thus, expediting decision-making processes. Satellite measure various spectral wavelengths of light reflected by plants (i.e., reflectance), and using satellite images, various vegetation indices (VI) can be calculated, such as the Normalized Difference Vegetation Index (NDVI), the Enhanced Vegetation Index (EVI), the Green Normalized Difference Vegetation Index (GNDVI), and the Spectral Feature Depth Vegetation Index (SFDVI). Each of these indices shows different vegetation properties via mathematical combinations of reflectance in two or more spectral wavelengths. For example, NDVI is calculated by reflectance in red and near infrared wavelengths, which is an effective VI to quantify vegetation greenness and vigor. This study aimed to compare the mentioned VI to evaluate their ability in estimating cereal rye (Secala cereale) mass and to compare them with a traditional method of biomass measurement, which relies on the standing height of the forage. From May 2 to 23, 2023, forage mass was measured in 18 paddocks (1.6 ha/paddock). Mass in the paddock was estimated by clipping four randomly located frames (0.49 m2) to ground level. The standing height of the forage was collected from 20 random points, measured in 0.625 cm increments, within each paddock. The forage mass of the paddocks measured from the clipped forage ranged from 212 to 3,382 kg/ha. PlanetScope satellite images from the day of forage mass sampling were obtained using the Planet Explorer platform. This satellite data was chosen due to its high spatial resolution (3 x 3 m2/pixel) and highly available data (theoretically daily). The four VI were calculated from images in QGIS software. A linear regression of each index against the forage mass estimated from the average of the clipped forage, was conducted in SAS using PROC REG. There was a significant (P < 0.01) correlation between each VI with the forage mass. The adjusted R2 was 0.60, 0.58, 0.41, 0.78 and 0.76 for NDVI, EVI, GNDVI, SFDVI and standing height, respectively. The root mean square error (RMSE) was 518, 528, 629, 379, and 396 for NDVI, EVI, GNDVI, SFDVI, and standing height, respectively. All VI exhibited moderately promising results in cereal rye mass estimation. In particular, SFDVI proved to be superior with the greatest adjusted R² and least RMSE, followed by the traditional method of estimating forage mass by standing height. This comparable result between the use of SFDVI from satellite images and traditional ground measurement indicates that the use of VI, SFDVI, specifically in our study, from satellite images can be an effective means for a fast cereal rye mass estimation.