Diagnostic Feed Values of Natural Grasslands Based on Multispectral Images Acquired by Small Unmanned Aerial Vehicle

Abstract

Grasslands are the largest renewable source of terrestrial chlorophytes. Furthermore, grasslands can be both fiber sources and the primary metabolizable energy source for ruminants. Therefore, rapid, accurate, and large-scale monitoring of grassland ecosystems is important to provide spatial information on forage quality control and rangeland management. In this experiment, 100 grassland sites were randomly selected in two study areas. A multiaxis unmanned aerial vehicle (UAV) made 26 flights over those areas to capture spectral images during August 2016, which enabled the acquisition of vegetation index values of the grassland sites. Next, grassland plots were harvested and the nutritional composition of the grass was determined. After selecting the most sensitive spectral information for each nutritional value, retrieval models for grassland nutrition were constructed. Predictor variables of the models were then tested on the samples. The results demonstrate that there are correlations between nutritional values and vegetation indices. The predicted values of the coefficient of determination (R2-P) and root mean square error (RMSE) for dry matter (DM) were 0.676% and 4.719%. The same values for crude protein (CP) were 0.653% and 1.361%. The R2-P and RMSE values for in vitro DM digestibility (IVDMD) prediction models were weak, but they could be improved by more sensitive wavelengths and improved mathematical models to fit the data. The results show that UAV remote sensing can be used to estimate the feed values of natural grassland and that this sensing approach provides a rapid, flexible, and efficient method of estimating feed values. Although the prediction models for nutritional values need to be improved, they still opened perspectives for the use of UAV-based remote sensing in rangeland management and grassland husbandry.