Contents lists available at ScienceDirect Geoderma journal homepage: www.elsevier.com/locate/geoderma Combination of Convolutional Neural Networks and Recurrent Neural Networks for predicting soil properties using Vis–NIR spectroscopy Jiechao Yang a,b , Xuelei Wang b, ⁎ , Ruihua Wang a,b , Huanjie Wang a,b a University of Chinese Academy of Sciences, School of Artificial Intelligence, Beijing 100049, China b Institute of Automation, Chinese Academy of Sciences, Beijing 100190, China ARTICLE INFO Handling Editor: Budiman Minasny Keywords: Vis–NIR spectroscopy Convolutional Neural Network Recurrent Neural Network Soil properties estimation Transfer learning ABSTRACT Visible and Near-infrared diffuse reflectance spectroscopy (Vis–NIR) serves as a rapid and non-destructive technique to estimate various soil properties. Recently, there is a growing need for developing a more accurate and robust calibration model in large soil spectral libraries to support the implementation of effective soil quality assessments and digital soil maps at national, continental and even global scales. Traditional calibration methods, such as partial least squares regression (PLSR), support vector machines regression(SVMR), multi- variate adaptive regression splines(MARS), random forests(RF), and artificial neural networks (ANN), may not be successfully applied in large spectral libraries due to their relatively weak generation performance in large regions. To overcome these weaknesses, we proposed a jointed Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) architecture called CCNVR, which combines the ability of CNN to extract the local and abstract features from the raw spectrum with the advantage of RNN to learn various dependencies of sequence features. We then compared the prediction accuracy of CCNVR with other conventional methods, namely, PLSR, SVMR, CNN, ANN, and RNN, on the selected soil properties of mineral soil samples in the Land Use/Land Cover Area Frame Survey (LUCAS) database. Of all calibration models, our proposed CCNVR achieved the best model performance with the lowest RMSE value (6.40, 0.45, 3.30, and 0.35 for OC, N, CEC, and pH, respectively) and the highest R 2 (0.73, 0.70, 0.73, and 0.86 for OC, N, CEC, and pH, respectively) for the selected properties, indicating the outstanding prediction ability of our proposed model. Besides, to quantify the ro- bustness of different calibration models, we added different levels of white noise on the original Vis–NIR spectra of the calibration set to observe how the prediction accuracy changes in the test set. The result showed that our proposed CCNVR model has a better resistance towards noise compared to other calibration models. Finally, we explored the transferability of our proposed CCNVR model. We extended the calibration model trained on the mineral samples to the organic samples through transfer learning. The result revealed that the transfer-based CCNVR fine-tuning model had a better prediction accuracy than that of the non-transfer CCNVR model with an improvement of R2 value from 0.79 to 0.84. The result demonstrated the excellent transferability of our pro- posed CCNVR model across different soil types and sample sizes. 1. Introduction The accurate estimation of soil properties is the basis of soil quality assessment and precision management. Conventional soil property de- termination methods require complex chemical laboratory analysis and specialized domain knowledge, which can be expensive and time-con- suming (Ben-Dor and Banin, 1995; Shi et al., 2015). Therefore, the Visible and Near-infrared diffuse reflectance (Vis–NIR) spectroscopy technique (Rossel et al., 2006) serves as an alternative method for evaluating different soil properties due to its simplicity, rapidity, and the fact that little or no sample preparation is required (Wang et al., 2015). However, the main challenge of Vis–NIR spectroscopy analysis is that the relationship between Vis–NIR spectra and corresponding soil properties is highly nonlinear, while there exists weak absorption in- tensity, the largely non-specific and severe overlap of absorption bands in the Vis–NIR region, which further complicates the Vis–NIR spectral analysis (Yuanyuan and Zhibin, 2018). Therefore, it is necessary to establish an accurate and reliable calibration model based on Vis–NIR spectroscopy to estimate soil nutrient content. Various calibration methods have been employed to establish the https://doi.org/10.1016/j.geoderma.2020.114616 Received 18 January 2020; Received in revised form 21 July 2020; Accepted 23 July 2020 ⁎ Corresponding author at: Institute of Automation, Chinese Academy of Sciences, 95 East Zhongguancun Road Haidian District, Beijing 100190, China. E-mail addresses: yangjiechao2018@ia.ac.cn (J. Yang), xuelei.wang@ia.ac.cn (X. Wang), wangruihua2019@ia.ac.cn (R. Wang), wanghuanjie2018@ia.ac.cn (H. Wang). Geoderma 380 (2020) 114616 0016-7061/ © 2020 Elsevier B.V. All rights reserved. T