Development of pedotransfer functions by machine learning for prediction of soil electrical conductivity and organic carbon content
- Benke, Kurt, Norng, Sorn, Robinson, Nathan, Chia, K., Rees, David, Hopley, J.
- Authors: Benke, Kurt , Norng, Sorn , Robinson, Nathan , Chia, K. , Rees, David , Hopley, J.
- Date: 2020
- Type: Text , Journal article
- Relation: Geoderma Vol. 366, no. (2020), p.
- Full Text:
- Reviewed:
- Description: The pedotransfer function is a mathematical model used to convert direct soil measurements into known and unknown soil properties. It provides information for modelling and simulation in soil research, hydrology, environmental science and climate change impacts, including investigating the carbon cycle and the exchange of carbon between soils and the atmosphere to support carbon farming. In particular, the pedotransfer function can provide input parameters for landscape design, soil quality assessment and economic optimisation. The objective of the study was to investigate the feasibility of using a generalised pedotransfer function derived with a machine learning method to predict soil electrical conductivity (EC) and soil organic carbon content (OC) for different regional locations in the state of Victoria, Australia. This strategy supports a unified approach to the interpolation and population of a single regional soils database, in contrast to a range of pedotransfer functions derived from local databases with measurement sets that may have limited transferability. The pedotransfer function generation was based on a machine learning algorithm incorporating the Generalized Linear Mixed Model with interactions and nested terms, with Residual Maximum Likelihood estimation, and a predictor-frequency ranking system with step-wise reduction of predictors to evaluate the predictive errors in reduced models. The source of the data was the Victorian Soil Information System (VSIS), which is a database administered for soil information and mapping purposes. The database contains soil measurements and information from locations across Victoria and is a repository of historical data, including monitoring studies. In total, data from 93 projects were available for inputs to modelling and analysis, with 5158 samples used to derive predictors for EC and 1954 samples used to derive predictors for OC. Over 500 models were tested by systematically reducing the number of predictors from the full model. Five-fold cross-validation was used for estimation of model mean-squared prediction error (MSPE) and mean-absolute percentage error (MAPE). The results were statistically significant with only a gradual reduction in error for the top-ranked 50 models. The prediction errors (MSPE and MAPE) of the top ranked model for EC are 0.686 and 0.635, and 0.413 and 0.474 for OC respectively. The four most frequently occurring predictors both for EC and OC prediction across the full set of models were found to be soil depth, pH, particle size distribution and geomorphological mapping unit. The possible advantages and disadvantages of this approach were discussed with respect to other machine learning approaches. © 2020 Elsevier B.V.
- Authors: Benke, Kurt , Norng, Sorn , Robinson, Nathan , Chia, K. , Rees, David , Hopley, J.
- Date: 2020
- Type: Text , Journal article
- Relation: Geoderma Vol. 366, no. (2020), p.
- Full Text:
- Reviewed:
- Description: The pedotransfer function is a mathematical model used to convert direct soil measurements into known and unknown soil properties. It provides information for modelling and simulation in soil research, hydrology, environmental science and climate change impacts, including investigating the carbon cycle and the exchange of carbon between soils and the atmosphere to support carbon farming. In particular, the pedotransfer function can provide input parameters for landscape design, soil quality assessment and economic optimisation. The objective of the study was to investigate the feasibility of using a generalised pedotransfer function derived with a machine learning method to predict soil electrical conductivity (EC) and soil organic carbon content (OC) for different regional locations in the state of Victoria, Australia. This strategy supports a unified approach to the interpolation and population of a single regional soils database, in contrast to a range of pedotransfer functions derived from local databases with measurement sets that may have limited transferability. The pedotransfer function generation was based on a machine learning algorithm incorporating the Generalized Linear Mixed Model with interactions and nested terms, with Residual Maximum Likelihood estimation, and a predictor-frequency ranking system with step-wise reduction of predictors to evaluate the predictive errors in reduced models. The source of the data was the Victorian Soil Information System (VSIS), which is a database administered for soil information and mapping purposes. The database contains soil measurements and information from locations across Victoria and is a repository of historical data, including monitoring studies. In total, data from 93 projects were available for inputs to modelling and analysis, with 5158 samples used to derive predictors for EC and 1954 samples used to derive predictors for OC. Over 500 models were tested by systematically reducing the number of predictors from the full model. Five-fold cross-validation was used for estimation of model mean-squared prediction error (MSPE) and mean-absolute percentage error (MAPE). The results were statistically significant with only a gradual reduction in error for the top-ranked 50 models. The prediction errors (MSPE and MAPE) of the top ranked model for EC are 0.686 and 0.635, and 0.413 and 0.474 for OC respectively. The four most frequently occurring predictors both for EC and OC prediction across the full set of models were found to be soil depth, pH, particle size distribution and geomorphological mapping unit. The possible advantages and disadvantages of this approach were discussed with respect to other machine learning approaches. © 2020 Elsevier B.V.
A digital soil map of Victoria-VicDSMv1
- Hopley, J., Rees, David, MaEwan, Richard, Clark, R., Benke, Kurt, Imhof, Mark, Robinson, Nathan, Bardos, David
- Authors: Hopley, J. , Rees, David , MaEwan, Richard , Clark, R. , Benke, Kurt , Imhof, Mark , Robinson, Nathan , Bardos, David
- Date: 2014
- Type: Text , Conference paper
- Relation: GlobalSoilMap: Basis of the Global Spatial Soil Information System - Proceedings of the 1st GlobalSoilMap Conference p. 185-189
- Full Text: false
- Reviewed:
- Description: This paper describes the production of the first version of a digital soil map for Victoria (VicDSMv1) which has combined existing soil point and polygon data. Four soil properties: pH, EC (electrical conductivity), clay percentage and Soil Organic Carbon (SOC) at the 6 depths specified by GlobalSoilMap.net have been mapped. The mapping has utilised data from 5,233 legacy sites collated from soil and land surveys conducted across Victoria since the 1930s and stored in the Victorian Soil Information System (VSIS). These sites were prepared by allocating property values to each of the 6 depths using equal area splines or depth weighted values. A land unit map for Victoria was derived from an overlay of map units from 32 surveys mapped at 100,000 scale or better. A dominant soil type at Suborder level in the Australian Soil Classification system (ASC) was assigned to each land unit. For each polygon a hierarchical grouping of sites from the VSIS was created using soil classification and location in relation to the polygon. A set of statistics for each soil property value at each set depth were calculated from the best available site cluster for each polygon. Metadata relating to property calculations have been collected. Creation of the VicDSMv1 has involved the preparation and entry of a large volume of legacy soil information into the VSIS. Consultation with current and retired soil surveyors during the process has enabled valuable expert knowledge to be captured into digital soil mapping. © 2014 Taylor & Francis Group, London, UK.
- Robinson, Nathan, Benke, Kurt, Hopley, J., MacEwan, Richard, Clark, R., Rees, David, Kitching, Matt, Imhof, Mark, Bardos, David
- Authors: Robinson, Nathan , Benke, Kurt , Hopley, J. , MacEwan, Richard , Clark, R. , Rees, David , Kitching, Matt , Imhof, Mark , Bardos, David
- Date: 2014
- Type: Text , Conference paper
- Relation: GlobalSoilMap: Basis of the Global Spatial Soil Information System - Proceedings of the 1st GlobalSoilMap Conference p. 353-358
- Full Text: false
- Reviewed:
- Description: Production of a digital soil map for the state of Victoria in Australia is subject to various errors arising from the use of legacy data (as is often the case around the globe). Potential sources of uncertainty for inputs and methods undertaken in the creation of a Victorian DSM version 1.0 (VicDSMv1) are identified. These sources of uncertainty are recognised and issues discussed including their potential contribution to error propagation. Examples include possible errors associated with legacy soil maps, soil sites, laboratory analysis and predictive modelling by regression or spline approaches. Experiences in processing of legacy data in Victoria are described and some aspects of incorporating uncertainty in data discussed. As part of this initial DSM exercise these uncertainties and contextual information will be captured as associated metadata. A framework, as a five component process model for integrated assessment of uncertainty, is suggested based on uncertainty in the mapping process. © 2014 Taylor & Francis Group, London, UK.
- «
- ‹
- 1
- ›
- »