A framework for optimizing environmental covariates to support model interpretability in digital soil mapping
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
A common practice in digital soil mapping (DSM) is to incorporate many environmental covariates into a machine-learning algorithm to predict the spatial patterns of soil attributes. Variance inflation factor (VIF), principal component analysis (PCA), and recursive feature elimination (RFE) are three statistical methods that can be used to reduce the number of covariates. This study aims 1) to compare VIF and PCA approaches; 2) to identify an approach to determine the minimum number of covariates in DSM to ensure model parsimony using RFE after using VIF; and 3) to examine methods to interpret the impact of covariates on the variability of the predicted soil properties. The study area was the province of British Columbia (BC), Canada. This study used legacy data for four soil properties to make digital soil maps: soil organic carbon (SOC%), pH, clay%, and coarse fragment (CF%). Seven models were made for each soil property to determine the influence on validation results by using a different number of covariates produced by various methods on validation results. The results showed that the number of covariates could be reduced from 70 to 4 to 12 with only a little or no difference in concordance correlation coefficient (CCC) validation results. The CCC results of pH models using 70 and 7 covariates were both 0.74, and for other soil properties, this difference was negligible. The validation results obtained from PCA models showed that the performance of PCA in reducing the number of covariates was not as effective as when using VIF. Moreover, this study showed that covariates related to precipitation were the most important for modeling SOC%, soil pH, and clay%. Topographic covariates were the most influential covariates for modeling soil CF%. This study emphasizes the potential benefits of combining various data reduction methods to achieve optimal outcomes and generate the most parsimonious and interpretable models.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it