Hyperspectral analysis of mangrove foliar chemistry using PLSR and support vector regression

Abstract

Hyperspectral remote sensing enables the large-scale mapping of canopy biochemical properties. This study explored the possibility of retrieving the concentration of nitrogen, phosphorus, potassium, calcium, magnesium, and sodium from mangroves in the Berau Delta, Indonesia. The objectives of the study were to (1) assess the accuracy of foliar chemistry retrieval, (2) compare the performance of models based on support vector regression (SVR), i.e. ε-SVR, ν-SVR, and least squares SVR (LS-SVR), to models based on partial least squares regression (PLSR), and (3) investigate which spectral transformations are best suited. The results indicated that nitrogen could be successfully modelled at the landscape level (R 2 = 0.67, root mean square error (RMSE) = 0.17, normalized RMSE (nRMSE) = 15%), whereas estimations of P, K, Ca, Mg, and Na were less encouraging. The developed nitrogen model was applied over the study area to generate a map of foliar N variation, which can be used for studying ecosystem processes in mangroves. While PLSR attained good results directly using all untransformed bands, the highest accuracy for nitrogen modelling was achieved using a combination of LS-SVR and continuum-removed derivative reflectance. All SVR techniques suffered from multicollinearity when using the full spectrum, and the number of independent variables had to be reduced by singling out the most informative wavelength bands. This was achieved by interpreting and visualizing the structure of the PLSR and SVR models.

Exploring the Potential of WorldView-2 Red-Edge Band-Based Vegetation Indices for Estimation of Mangrove Leaf Area Index with Machine Learning Algorithms

“…The variable importance measure was based on the inner product of the normalized predictor variables and the α-vector proposed by Üstün et al [75]. These positive or negative values were similar to regression coefficients, while the absolute values were the measure of variable importance [68].…”

“…The sensitivities of input variables were determined by quantifying variables' importance using all three methods. The importance measures were standardized to a standard deviation [68].…”

Limited Effects of Water Absorption on Reducing the Accuracy of Leaf Nitrogen Estimation

“…The absorption features of nitrogen, which mainly result from the vibration of chemical bonds, are shadowed by water or are affected by the signal/noise ratio (SNR) [38]. Therefore, many studies at the canopy scale removed spectral regions with strong water absorptions after atmospheric correction, e.g., [17,29,30,32,39,40]. However, this practice may lead to the loss of useful spectral information.…”

Assessing the Suitability of Future Multi- and Hyperspectral Satellite Systems for Mapping the Spatial Distribution of Norway Spruce Timber Volume

“…If the emphasis is on predicting the responses (timber volume) and not necessarily on trying to understand the underlying relationships between the spectral variables, PLSR is an appropriate choice. Developed in the 1960s by the Swedish statistician Herman Wold, PLSR has qualified as a standard tool in chemistry and engineering [51], and has been applied in numerous spectrometric studies at the laboratory, field, and canopy scale [52][53][54][55]. It is based on the assumption that the response variable is explained by only a few underlying or latent factors (PLS components) that account for most of the variation in the response.…”