Article
Soil Science
Songchao Chen, Jie Xue, Zhou Shi
Summary: Ensemble modelling (EM) is widely used in soil information prediction, and a new method called spectral-guided ensemble modelling (S-GEM) has been proposed to improve soil spectroscopic prediction. The results show that S-GEM outperforms EM in predicting soil properties using vis-NIR spectra. Therefore, S-GEM has a high potential to enhance the accuracy of soil spectroscopic prediction.
Article
Chemistry, Analytical
Zijin Bai, Modong Xie, Bifeng Hu, Defang Luo, Chang Wan, Jie Peng, Zhou Shi
Summary: This study evaluates the feasibility of different spectral bands variable selection methods for predicting soil organic carbon (SOC) content and analyzes the impact of different methods on the prediction accuracy. The results show that using the Competitive adaptive reweighted sampling (CARS) method to select feature bands, combined with convolutional neural network (CNN) modeling, can achieve higher accuracy in predicting SOC.
Article
Soil Science
Meihua Yang, Songchao Chen, Dongyun Xu, Yongsheng Hong, Shuo Li, Jie Peng, Wenjun Ji, Xi Guo, Xiaomin Zhao, Zhou Shi
Summary: The large-scale soil spectral library (SSL) provides abundant information for predicting soil properties, but using SSL for predicting soil information from in situ spectra is still a challenge. This study compared different strategies for predicting soil organic matter (SOM) using SSL and found that the mean squared Euclidean distance (msd) is an optimal indicator for selecting representative samples. The recommended strategy depends on the availability of in situ and dry spectra. These findings contribute to efficient SOM prediction in situ by integrating large-scale SSL.
Article
Geography, Physical
Yilin Bao, Fengmei Yao, Xiangtian Meng, Jiahua Zhang, Huanjun Liu, Abdul Mounem Mouazen
Summary: This study proposes a mapping framework for soil organic carbon (SOC), which integrates satellite earth observations and proximal sensing spectral data to reconstruct high-resolution images. The clustered probability analysis is applied to develop an accurate SOC prediction model at different scales. The results demonstrate that the reconstructed images outperform the individual spectral data and Sentinel-2A models, and the clustering probability model achieves better performance in quantifying SOC at the regional scale.
ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING
(2023)
Article
Agriculture, Multidisciplinary
Camilo Ernesto Bohorquez-Sanchez, Saulo Augusto Quassi de Castro, Joao Luis Nunes Carvalho, Sarah Tenelli, Risely Ferraz-Almeida, Renata Alcarde Sermarini, Izaias Pinheiro Lisboa, Rafael Otto
Summary: Large amounts of straw remained in the sugarcane fields after harvest. Indiscriminate straw removal or high amounts of straw on the field can negatively affect soil functions and crop yields. Leguminosae species grown during the sugarcane renewal period can potentially offset the negative implications of unsustainable straw management on soil and crop yield.
AGRICULTURE ECOSYSTEMS & ENVIRONMENT
(2023)
Article
Geosciences, Multidisciplinary
Maryem Arshad, Dongxue Zhao, Ehsan Zare, Michael Sefton, John Triantafilis
Summary: The study utilized stepwise SVM and calibration dataset to determine the optimal digital data, showing that combining digital data resulted in better accuracy compared to individual gamma-ray or ECa data. In independent validation, results from Mossman reflected the general rank order of different approaches, with site-specific predictions performing the best.
Article
Engineering, Electrical & Electronic
Wudi Zhao, Zhilu Wu, Zhendong Yin, Dasen Li
Summary: In this article, a new deep-learning-based SMC influence removal network (MIRNet) is proposed to eliminate the interference of soil moisture content when estimating soil organic carbon using visible and near-infrared spectra measured in situ. The MIRNet utilizes a spectral extraction module and a context extraction module to extract soil spectral characteristics and context information, respectively. The experimental results show that MIRNet achieves competitive results compared with existing methods.
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
(2023)
Article
Soil Science
Michael Vohland, Bernard Ludwig, Michael Seidel, Christopher Hutengs
Summary: The study compared laboratory-based and in situ fusion approaches for visible to near and mid-infrared chemometric modelling in soil studies. It found that regionally stratified approach and model ensemble averaging were beneficial for accurate estimations. The research suggested that simple averaging procedures could advance multi-sensor applications integrating vis-NIR and MIR data for in situ soil spectroscopy, especially in regions with heterogeneous soil conditions.
Article
Soil Science
Nan Li, Dongxue Zhao, Maryem Arshad, Michael Sefton, John Triantafilis
Summary: This paper explores the application of digital soil mapping (DSM) in sugarcane growing areas and compares different models and data. The results show that using the Cubist-RK model combined with digital data can provide more accurate predictions of exchangeable sodium percentage (ESP) in the soil, while the clustering of digital data is effective for delineating management zones.
SOIL USE AND MANAGEMENT
(2022)
Article
Agriculture, Multidisciplinary
Dongxue Zhao, Maryem Arshad, Jie Wang, John Triantafilis
Summary: The study aims to determine the best model for predicting topsoil exchangeable calcium, magnesium, potassium, and sodium; evaluate the applicability of the best topsoil model for subsurface and subsoil exchangeable cations; explore the effect of spiking on subsurface and subsoil prediction using the topsoil spectral library; and assess if building a profile spectral library with all depths improves prediction. PLSR was superior for predicting topsoil exchangeable cations, while Cubist outperformed PLSR in certain cases when spiking was applied and a profile spectral library was considered. Topsoil PLSR could also be used to predict subsurface and subsoil exchangeable calcium and magnesium, with spiking improving prediction. Profile spectral library achieved equivalent results when considering topsoil samples coupled with spiking.
COMPUTERS AND ELECTRONICS IN AGRICULTURE
(2021)
Article
Geosciences, Multidisciplinary
Dongxue Zhao, Jie Wang, Xueyu Zhao, John Triantafilis
Summary: This study compared the importance of proximally and remotely sensed data for predicting clay at district scale, using different models and averaging techniques. The results showed that gamma-ray data was crucial for topsoil clay prediction, while slope was important for subsoil. Random forest model was found to be the best for predicting topsoil clay, with Granger-Ramanathan averaging recommended as a protocol for district-scale clay prediction.
Article
Agriculture, Multidisciplinary
Jie Wang, Dongxue Zhao, Ehsan Zare, Michael Sefton, John Triantafilis
Summary: This study explores various models to predict soil organic carbon, and finds that the hybrid models RFRK and SVMRK perform the best in terms of prediction accuracy. The DSM using RFRK provides a method for farmers to determine nitrogen fertilizer rates based on soil conditions, resulting in a potential decrease in fertilizer application cost.
COMPUTERS AND ELECTRONICS IN AGRICULTURE
(2022)
Article
Geosciences, Multidisciplinary
Linghua Duo, Yanan Li, Ming Zhang, Yuxi Zhao, Zhenhua Wu, Dongxue Zhao
Summary: This study explores the complex dynamic evolution mechanism of urban ecosystem resilience based on the three resilience characteristics. Nanchang is selected as the assessment target. The results show that the trend of land-use change in Nanchang is characterized by the decrease in arable land and ecological land and the increase in construction land. The overall ecosystem resilience of Nanchang shows a gradual decline and spatial heterogeneity.
FRONTIERS IN EARTH SCIENCE
(2022)
Article
Ecology
Peng Zhou, Dongxue Zhao, Xiao Liu, Linghua Duo, Bao-Jie He
Summary: This study analyzed the dynamic vegetation changes in Alxa League, China by taking into account various factors such as land cover types, elevation, climate, and air quality. The results showed a significant drop in vegetation index after 2012, which slowly recovered after 2015. High vegetation index values were found in urban areas with frequent human activities. The northwest region exhibited a slight degradation trend, while the southeast region showed a slight improvement trend. The vegetation index was negatively correlated with temperature, precipitation, soil moisture, and total evaporation, and positively correlated with precipitation and soil moisture.
FRONTIERS IN ECOLOGY AND EVOLUTION
(2022)
Article
Agriculture, Multidisciplinary
Dongxue Zhao, Joseph X. Eyre, Erin Wilkus, Peter de Voil, Ian Broad, Daniel Rodriguez
Summary: Most field crop phenotyping research has focused on above-ground parts of crops, neglecting the rooting system and its activity. In this study, a new approach using electromagnetic induction instrument and crop canopy sensing technologies was proposed and tested to characterize crop water use and root activity. The results showed that this approach could accurately predict 3D crop water use and capture complex genotype-environment-management dynamics.
COMPUTERS AND ELECTRONICS IN AGRICULTURE
(2022)
Article
Soil Science
Jie Wang, Xueyu Zhao, Kirstin E. Deuss, David R. Cohen, John Triantafilis
Summary: Soil cation exchange capacity (CEC) and pH have an impact on soil condition. Sugar Research Australia introduced nutrient guidelines based on CEC and pH of topsoil to improve soil capability in sugarcane growing areas. A three-dimensional digital soil mapping framework was used to predict CEC and pH using various data sources. The results can guide lime application in the sugarcane fields.
Article
Soil Science
Xueyu Zhao, Jie Wang, Dongxue Zhao, John Triantafilis
Summary: This study compared different digital soil mapping models to predict soil organic carbon content in the sugarcane industry. The results showed that hybrid models had higher predictive accuracy compared to non-hybrid models. Using a hybrid model of MLR with 75 calibration samples for digital soil mapping resulted in a significant cost reduction and consistent results compared to the industry average.
NUTRIENT CYCLING IN AGROECOSYSTEMS
(2023)
Article
Soil Science
Evangeline Fung, Jie Wang, Xueyu Zhao, Mohammad Farzamian, Barry Allred, William Bruce Clevenger, Philip Levenson, John Triantafilis
Summary: Cation exchange capacity (CEC) is the ability of soil to hold exchangeable cations, including potassium (K). This study focuses on developing a linear regression between soil electrical conductivity (ECa) and measured CEC, as well as between estimated sigma and CEC. The research also investigates the minimum number of calibration sample sites needed for accurate prediction, and proposes methods to improve predictive models and reduce confidence intervals.
SOIL & TILLAGE RESEARCH
(2023)
Article
Geochemistry & Geophysics
Xueyu Zhao, Jie Wang, Dongxue Zhao, Michael Sefton, John Triantafilis
Summary: The study explores the use of electromagnetic induction instruments to predict soil cation exchange capacity (CEC) in sugarcane growing soil. By measuring soil apparent electrical conductivity (ECa), it is possible to estimate CEC, and combining this with depth and location data improves prediction accuracy. The results show a moderate correlation between estimated electrical conductivity (a) and CEC. These findings have implications for digitally mapping topsoil CEC and applying lime in the Australian sugarcane industry.
JOURNAL OF ENVIRONMENTAL AND ENGINEERING GEOPHYSICS
(2022)