Integrating Airborne Hyperspectral, Topographic, and Soil Data for Estimating Pasture Quality Using Recursive Feature Elimination with Random Forest Regression

Abstract

Accurate and efficient monitoring of pasture quality on hill country farm systems is crucial for pasture management and optimizing production. Hyperspectral imaging is a promising tool for mapping a wide range of biophysical and biochemical properties of vegetation from leaf to canopy scale. In this study, the potential of high spatial resolution and airborne hyperspectral imaging for predicting crude protein (CP) and metabolizable energy (ME) in heterogeneous hill country farm was investigated. Regression models were developed between measured pasture quality values and hyperspectral data using random forest regression (RF). The results proved that pasture quality could be predicted with hyperspectral data alone; however, accuracy was improved after combining the hyperspectral data with environmental data (elevation, slope angle, slope aspect, and soil type) where the prediction accuracy for CP was R 2 CV (cross-validated coefficient of determination) = 0.70, RMSE CV (cross-validated root mean square error) = 2.06%, RPD CV (cross-validated ratio to prediction deviation) = 1.82 and ME: R 2 CV = 0.75, RMSE CV = 0.65 MJ/kg DM, RPD CV = 2.11. Interestingly, the accuracy was further out-performed by considering important hyperspectral and environmental variables using RF combined with recursive feature elimination (RFE) (CP: R 2 CV = 0.80, RMSE CV = 1.68%, RPD CV = 2.23; ME: R 2 CV = 0.78, RMSE CV = 0.61 MJ/kg DM, RPD CV = 2.19). Similar performance trends were noticed with validation data. Utilizing the best model, spatial pasture quality maps were created across the farm. Overall, this study showed the potential of airborne hyperspectral data for producing accurate pasture quality maps, which will help farm managers to optimize decisions to improve environmental and economic benefits.

Predicting Forage Quality of Grasslands Using UAV-Borne Imaging Spectroscopy

“…However, PLSR models resulted in the lowest accuracies both for CP and ADF in this study. Moreover, Pullanagari et al [34] achieved an nRMSE of 11.2% for CP with the RFR model, and Singh et al [33] obtained an nRMSE of 21.7% for ADF with the same modelling algorithm. It is noteworthy that the studies mentioned above only tested one predictive modelling algorithm; thus, no conclusions are possible considering the comparison with other algorithms.…”

“…The most frequently applied statistical modelling method is the linear regression (simple linear, step-wise linear) with selected highly correlated spectral features, such as wavebands [20,27], normalised difference spectral indices (NDSIs) [18,24,28], spectral ratios (SRs) [15,29], and other well-known vegetation indices (e.g., NDVI, SAVI, NDRE) [17,23]. Predictive modelling (also known as machine learning) algorithms, such as partial least squares regression (PLSR) [16,27,[30][31][32], random forest regression (RFR) [24,33,34], and artificial neural network [20,21,35], were employed to estimate forage quality parameters using highly correlated spectral reflectance data. Predictive modelling algorithms frequently enhanced the predictive capability compared with the simpler linear regression models [24,34].…”

“…Predictive modelling (also known as machine learning) algorithms, such as partial least squares regression (PLSR) [16,27,[30][31][32], random forest regression (RFR) [24,33,34], and artificial neural network [20,21,35], were employed to estimate forage quality parameters using highly correlated spectral reflectance data. Predictive modelling algorithms frequently enhanced the predictive capability compared with the simpler linear regression models [24,34]. Moreover, there are several other algorithms in predictive modelling, namely, Remote Sens.…”

Estimating Forest Stock Volume in Hunan Province, China, by Integrating In Situ Plot Data, Sentinel-2 Images, and Linear and Machine Learning Regression Models

“…The RF is a decision tree algorithm and an effective machine learning model for predicting a forest of variables. Based on its powerful modeling capabilities, the RF regression has been widely used in scientific research [94][95][96][97][98][99]. The principle of the RF algorithm is to use the bootstrap method to randomly extract multiple samples to generate a group of regression trees (ntree) from the original sample population.…”

Recursive Feature Elimination and Random Forest Classification of Meadows and Dry Grasslands in Lowland River Valleys of Poland Based on Airborne Hyperspectral and LiDAR Data Fusion

“…The Recursive Feature Elimination combined with Random Forest (RFE-RF) classification has been reported as a promising technique to effectively handling the fusion of diverse data sources, such as HS and LiDAR data, at the same time generating unbiased and stable classification results in different application fields [64][65][66][67][68]. One of the objectives of this paper was therefore to develop an RFE-RF system, capable of combining the potentiality of RF classification and RFE feature selection, at the same time automatizing the analysis of HS and ALS data for our habitats mapping purpose.…”