Combining Partial Least Squares and the Gradient-Boosting Method for Soil Property Retrieval Using Visible Near-Infrared Shortwave Infrared Spectra

Lanfa Liu,Manfred Buchroithner,Min Ji

doi:10.3390/rs9121299

Abstract

Soil spectroscopy has experienced a tremendous increase in soil property characterisation, and can be used not only in the laboratory but also from the space (imaging spectroscopy). Partial least squares (PLS) regression is one of the most common approaches for the calibration of soil properties using soil spectra. Besides functioning as a calibration method, PLS can also be used as a dimension reduction tool, which has scarcely been studied in soil spectroscopy. PLS components retained from high-dimensional spectral data can further be explored with the gradient-boosted decision tree (GBDT) method. Three soil sample categories were extracted from the Land Use/Land Cover Area Frame Survey (LUCAS) soil library according to the type of land cover (woodland, grassland, and cropland). First, PLS regression and GBDT were separately applied to build the spectroscopic models for soil organic carbon (OC), total nitrogen content (N), and clay for each soil category. Then, PLS-derived components were used as input variables for the GBDT model. The results demonstrate that the combined PLS-GBDT approach has better performance than PLS or GBDT alone. The relative important variables for soil property estimation revealed by the proposed method demonstrated that the PLS method is a useful dimension reduction tool for soil spectra to retain target-related information.

Highlights

Monitoring the status of soil is very important for tackling many challenges including food security, climate change, land degradation, and biodiversity [1]
To make a comparison with the following results obtained from Partial least squares (PLS)-Gradient-Boosted Decision Trees (GBDT), soil spectroscopic models for organic carbon (OC), N, and clay were developed using PLS regression with the same dataset (Figure 4)
Soil spectra measured in the laboratory typical have several hundred or even thousand bands, which would be a problem for the gradient-boosting model when directly using such high-dimensional data as inputs

Summary

Introduction

Monitoring the status of soil is very important for tackling many challenges including food security, climate change, land degradation, and biodiversity [1]. As a fast, cost-effective, and environmental-friendly analytical technique, has successfully been utilised to retrieve soil properties and has experienced a tremendous increase in the past years. It has been shown that soil spectra across the Visible Near-Infrared Shortwave Infrared (VIS–NIR–SWIR; 400–2500 nm) spectral region are characterised by significant spectral signals [3,4,5,6], which makes it possible for quantitative analysis of soil properties. The wide spread use of visible and infrared spectroscopy can resolve the trade-off between the growing need of large scale soil information and its high cost [7]. Using spectral measurements and corresponding soil properties measured by soil analyses, soil spectroscopy can be adopted to quantitatively estimate many soil properties, such as organic matter, heavy metals, clay content, exchangeable potassium, and electrical conductivity [8,9,10]

Objectives

Methods

Results

Conclusion