Selecting a proper set of covariates is one of the most important factors that influence the accuracy of digital soil mapping (DSM).  The statistical or machine learning methods for selecting DSM covariates are not available for those situations with limited samples.  To solve the problem, this paper proposed a case-based method which could formalize the covariate selection knowledge contained in practical DSM applications.  The proposed method trained Random Forest (RF) classifiers with DSM cases extracted from the practical DSM applications and then used the trained classifiers to determine whether each one potential covariate should be used in a new DSM application.  In this study, we took topographic covariates as examples of covariates and extracted 191 DSM cases from 56 peer-reviewed journal articles to evaluate the performance of the proposed case-based method by Leave-One-Out cross validation.  Compared with a novices’ commonly-used way of selecting DSM covariates, the proposed case-based method improved more than 30% accuracy according to three quantitative evaluation indices (i.e., recall, precision, and F1-score).  The proposed method could be also applied to selecting the proper set of covariates for other similar geographical modeling domains, such as landslide susceptibility mapping, and species distribution modeling.

 

In addition to soil samples, conventional soil maps, and experienced soil surveyors, text about soils (e.g., soil survey reports) is an important potential data source for extracting soil–environment relationships.  Considering that the words describing soil–environment relationships are often mixed with unrelated words, the first step is to extract the needed words and organize them in a structured way.  This paper applies natural language processing (NLP) techniques to automatically extract and structure information from soil survey reports regarding soil–environment relationships.  The method includes two steps: (1) construction of a knowledge frame and (2) information extraction using either a rule-based method or a statistic-based method for different types of information.  For uniformly written text information, the rule-based approach was used to extract information.  These types of variables include slope, elevation, accumulated temperature, annual mean temperature, annual precipitation, and frost-free period.  For information contained in text written in diverse styles, the statistic-based method was adopted.  These types of variables include landform and parent material.  The soil species of China soil survey reports were selected as the experimental dataset.  Precision (P), recall (R), and F₁-measure (F1) were used to evaluate the performances of the method.  For the rule-based method, the P values were 1, the R values were above 92%, and the F1 values were above 96% for all the involved variables.  For the method based on the conditional random fields (CRFs), the P, R and F1 values for the parent material were, respectively, 84.15, 83.13, and 83.64%; the values for landform were 88.33, 76.81, and 82.17%, respectively.  To explore the impact of text types on the performance of the CRFs-based method, CRFs models were trained and validated separately by the descriptive texts of soil types and typical profiles.  For parent material, the maximum F1 value for the descriptive text of soil types was 90.7%, while the maximum F1 value for the descriptive text of soil profiles was only 75%.  For landform, the maximum F1 value for the descriptive text of soil types was 85.33%, which was similar to that of the descriptive text of soil profiles (i.e., 85.71%).  These results suggest that NLP techniques are effective for the extraction and structuration of soil–environment relationship information from a text data source.

Conventional soil maps contain valuable knowledge on soil–environment relationships.  Such knowledge can be extracted for use when updating conventional soil maps with improved environmental data.  Existing methods take all polygons of the same map unit on a map as a whole to extract the soil–environment relationship.  Such approach ignores the difference in the environmental conditions represented by individual soil polygons of the same map unit.  This paper proposes a method of mining soil–environment relationships from individual soil polygons to update conventional soil maps.  The proposed method consists of three major steps.  Firstly, the soil–environment relationships represented by each individual polygon on a conventional soil map are extracted in the form of frequency distribution curves for the involved environmental covariates.  Secondly, for each environmental covariate, these frequency distribution curves from individual polygons of the same soil map unit are synthesized to form the overall soil–environment relationship for that soil map unit across the mapped area.  And lastly, the extracted soil–environment relationships are applied to updating the conventional soil map with new, improved environmental data by adopting a soil land inference model (SoLIM) framework.  This study applied the proposed method to updating a conventional soil map of the Raffelson watershed in La Crosse County, Wisconsin, United States.  The result from the proposed method was compared with that from the previous method of taking all polygons within the same soil map unit on a map as a whole.  Evaluation results with independent soil samples showed that the proposed method exhibited better performance and produced higher accuracy.