Estimation of paddy rice leaf area index using machine learning methods based on hyperspectral data from multi-year experiments

Abstract

The performance of three machine learning methods (support vector regression, random forests and artificial neural network) for estimating the LAI of paddy rice was evaluated in this study. Traditional univariate regression models involving narrowband NDVI with optimized band combinations as well as linear multivariate calibration partial least squares regression models were also evaluated for comparison. A four year field-collected dataset was used to test the robustness of LAI estimation models against temporal variation. The partial least squares regression and three machine learning methods were built on the raw hyperspectral reflectance and the first derivative separately. Two different rules were used to determine the models’ key parameters. The results showed that the combination of the red edge and NIR bands (766 nm and 830 nm) as well as the combination of SWIR bands (1114 nm and 1190 nm) were optimal for producing the narrowband NDVI. The models built on the first derivative spectra yielded more accurate results than the corresponding models built on the raw spectra. Properly selected model parameters resulted in comparable accuracy and robustness with the empirical optimal parameter and significantly reduced the model complexity. The machine learning methods were more accurate and robust than the VI methods and partial least squares regression. When validating the calibrated models against the standalone validation dataset, the VI method yielded a validation RMSE value of 1.17 for NDVI(766,830) and 1.01 for NDVI(1114,1190), while the best models for the partial least squares, support vector machine and artificial neural network methods yielded validation RMSE values of 0.84, 0.82, 0.67 and 0.84, respectively. The RF models built on the first derivative spectra with mtry = 10 showed the highest potential for estimating the LAI of paddy rice.

Effects of Growth Stage Development on Paddy Rice Leaf Area Index Prediction Models

“…These models also lack the ability to parameterize spatiotemporal variability [26]. PLSR has been considered to be a powerful alternative to univariate methods and provides better performance in most cases [27][28][29], although there is a study that reported the opposite results [22]. Moreover, the potential performance of the state-of-the-art machine learning methods, such as SVR, RF and ANN, has been explored in several studies [14,15,30].…”

The Value of Sentinel-2 Spectral Bands for the Assessment of Winter Wheat Growth and Development

“…Although simple to apply, the development of VIs are often time-, locationand scale-specific, (i.e., leaf or canopy-scale [18]). Furthermore, VIs make simplistic assumptions about the reflectance properties of a target and typically use only two to three fixed spectral bands [19], thus under-exploiting the potential of Earth observation multispectral sensors. Alternatively, machine learning approaches have the potential to generate adaptive, robust and non-linear relationships between all spectral bands and ground measurements [20].…”

Improved Winter Wheat Spatial Distribution Extraction from High-Resolution Remote Sensing Imagery Using Semantic Features and Statistical Analysis

“…For example, spectral indexes, which have been widely used in the classification of middle-and low-resolution remote sensing imagery, are obtained by statistical analysis of the spectral information of the pixels [8]. Commonly used methods include various vegetation indexes [9,10], the Automated Water Extraction Index (AWEI) [11], the Normalized Difference Built-up Index (NDBI) [12], and the Remote Sensing Ecological Index (RSEI) [13]. The Enhanced Vegetation Index (EVI) [8], Normalized Different Vegetation Index (NDVI) [10], and other indexes derived from NDVI are effective at extracting vegetation information and have been widely used for extracting crop spatial distributions from low-resolution remote sensing imagery.…”

“…Commonly used methods include various vegetation indexes [9,10], the Automated Water Extraction Index (AWEI) [11], the Normalized Difference Built-up Index (NDBI) [12], and the Remote Sensing Ecological Index (RSEI) [13]. The Enhanced Vegetation Index (EVI) [8], Normalized Different Vegetation Index (NDVI) [10], and other indexes derived from NDVI are effective at extracting vegetation information and have been widely used for extracting crop spatial distributions from low-resolution remote sensing imagery. Some researchers have taken advantage of the high temporal resolution of middle-and low-spatial resolution remote sensing imagery to obtain the spectral index characteristics of a time series before extracting crop information with good results [14][15][16].…”