Abstract
Quantifying the effects of cultivation on soil properties is a persistent challenge in soil science, primarily due to the inherent heterogeneity of landscapes and the lack of uncultivated "native" baselines. This study addresses these issues by comparing forested and cultivated deep loess soils in Missouri using a three-dimensional grid sampling approach. By applying Multiple Regression and Response Surface Methodology (RSM), researchers modeled the horizontal and vertical distributions of soil properties—specifically horizon thicknesses and clay distribution—across hillslopes. The analysis revealed that second- and third-order response surface equations effectively characterize soil variability as a function of slope, distance from the summit, and land management. The results demonstrate that cultivation significantly alters the mathematical "response surface" of soil properties, providing a systematic framework for comparing native and managed landscapes.
Introduction
Soil scientists and agronomists have long struggled to account for the natural variability of soil across a landscape. Traditional experimental designs often rely on "blocking" and Analysis of Variance (ANOVA) to manage this heterogeneity. However, these methods frequently fail to capture the continuous nature of soil properties across sloping terrain and offer limited predictive power.
To accurately quantify the impact of human-induced erosion and long-term cultivation, a robust baseline is required. This study utilizes native forested areas as a control to compare against cultivated sites. The objective is to move beyond simple mean comparisons toward predictive models that relate visible landscape features—such as topography and slope position—to subsurface soil characteristics.
Methodology
Sampling Design
The researchers implemented a 3-dimensional grid sampling strategy across six hillslopes in Missouri. Sampling extended from the hillslope summit down to the lower backslope. The study focused on four primary dependent variables:
- Depth to maximum clay percentage.
- Thickness of the "A" horizon (topsoil).
- Thickness of the subsurface horizon.
- Depth to the top of the "B" horizon (subsoil).
Statistical Approach: Response Surface Methodology
Rather than assuming simple linear relationships, the study employed Response Surface Methodology (RSM). This approach treats soil properties as a continuous surface influenced by spatial coordinates and management factors. The independent variables included percent slope ($X_1$), distance from the summit ($X_2$), and site-specific indicators ($X_3$ through $X_7$).
Through Multiple Linear Regression, the researchers determined the appropriate polynomial order for each property:
- Second-order models (including squared terms and interactions) were sufficient for modeling clay depth and "A" horizon thickness.
- Third-order models (including cubic terms) were necessary to capture the complexity of subsurface thickness and the depth to the "B" horizon.
The General Linear Test (GLT) was used to verify if the inclusion of complex terms or management variables (forest vs. cultivation) significantly improved the model's predictive accuracy.
Study Area and Datasets
The research focused on deep loess (wind-blown silt) soils in two Missouri counties adjacent to the Missouri River:
- Saline County: Four sites mapped as the Knox series (two forested, two cultivated since the early 1900s).
- Boone County: Two sites mapped as the Menfro series (one forested, one cultivated).
The forested sites provided the necessary "native" baseline, while the cultivated sites represented nearly a century of agricultural management.
Key Insights and Results
Topography as a Systematic Driver
The study confirmed that landscape features like slope and distance from the summit are not merely sources of statistical "noise." Instead, they are systematic drivers of soil properties. The regression models explained a substantial portion of the variance, with $R^2$ values as follows:
- 0.76 for Depth to Top of "B" horizon.
- 0.73 for Subsurface Horizon Thickness.
- 0.53 for Depth to Clay Max.
- 0.30 for "A" Horizon Thickness.
The Impact of Cultivation
Land management does more than just alter the average thickness of a soil layer; it fundamentally changes the mathematical relationship between the soil and the landscape. Cultivated sites exhibited significantly different response surfaces compared to forested sites.
One of the most striking findings was the "Plow Layer" Effect. In cultivated soils, the "A" horizon often incorporates material from lower horizons (E or B) due to the combined effects of tillage and erosion. Consequently, the "A" horizon in farmed soil becomes a reflection of management history rather than a natural functional layer. Furthermore, all cultivated sites showed a marked decrease in the depth to the "B" horizon from the summit to the shoulder, indicating accelerated erosion and "truncation" of the soil profile.
Spatial Erosion Patterns
The analysis revealed that the thinning of the "A" horizon was twice as aggressive from the shoulder to the backslope as it was from the summit to the shoulder. This identifies the shoulder-to-backslope transition as the highest-risk zone for soil loss in these loess landscapes.
Conclusion
By integrating analysis of variance within a multiple regression framework, this study provides a systematic method for describing and comparing soil properties across different land uses. While the models are empirical and specific to the deep loess soils of Missouri, they demonstrate that topography and management can be used to predict soil profile characteristics with significant accuracy. Future work may improve these models by incorporating additional factors of soil formation, such as granular climatic data and biological activity, to further reduce the amount of undescribed soil variability.
Summary updated