Calibration and Validation of a Soil‐Landscape Model for Predicting Soil Drainage Class

Abstract

The soil‐landscape relationships frequently used for soil mapping activities are seldom documented. We developed a statistical model that relates soil drainage classes to eight landscape parameters describing slope morphology, proximity to surface drainage features, and soil parent material. Soil profiles and landscape parameters were described at 305 randomly selected sampling points within the Mifflintown 7.5‐min topographic quadrangle in the unglaciated ridge and valley physiographic province of central Pennsylvania. Variables defining the spatial structure of the landscape were derived from digitized 1:24 000 scale maps of streams and drainageways, surficial geology, bedrock geology, and a digital elevation model. These data were stored in a geographic information system and overlaid with the sampling point locations to define a database of 305 known soil‐landscape combinations. These soil‐landscape combinations were used to derive a statistical soil‐landscape model using multivariate discriminant analysis and class frequency information. For soil drainage class, a 74% overall agreement with field observations was found for the model using a cross‐validation approach, compared with 69% for the published soil survey. The model correctly predicted a majority of the observations within each drainage class and provided a consistent method of extrapolating point information about soils to the three‐dimensional landscape.

Australia-wide predictions of soil properties using decision trees

“…Much of the early development of the soil factor equation can be attributed to Jenny and Leonard (1934) and Jenny (1941) who sought to validate the soil factor equation by relating changes in soil properties to changes in climate across a WestEast transect in the United States. Since then there has been a large body of literature that relates soil properties to landscape characteristics (e.g., Walker et al, 1968;Kreznor et al, 1989;Bell et al, 1992;Odeh et al, 1994). Computer technology has made environmental variables more readily available and facilitated a greater pursuit of quantitative relationships with soil properties.…”

Soil–landscape modeling across a physiographic region: Topographic patterns and model transportability

“…Soil organic C (SOC) storage and dynamics have frequently been the focus of soil-landscape modeling efforts (Arrouays et al, 1995(Arrouays et al, , 1998Bell et al, 2000;Chaplot et al, 2001;Abnee et al, 2004a,b;Terra et al, 2004;Thompson and Kolka, 2005). Hydric and hydromorphic soil delineation (Thompson et al, 1997;Chaplot et al, 2000), net primary productivity (Gessler et al, 2000), soil drainage class (Bell et al, 1992, and soil water dynamics (Chamran et al, 2002) have also been successfully modeled.…”

“…Several authors (Bell et al, 1992Thompson et al, 1997;Chaplot et al, 2000;Thompson and Kolka, 2005) have employed an independent data set (collected from the same study site and not used in model development) to validate their models. Chaplot et al (2003) have provided the most rigorous test of model transportability across a region.…”

Model prediction of soil drainage classes based on digital elevation model parameters and soil attributes from coarse resolution soil maps

“…Soil toposequences were developed to describe the way water moves through and over landscapes (Moore et al 1993;Thompson et al 1997Thompson et al , 1998Campling et al 2002). Bell et al (1992) developed a soil-landscape model to incorporate a variety of topographical and parent material data to predict soil drainage distributions in south-central Pennsylvania. The results showed that soil parent material was an important factor that explained 74% of the field observations when soil was classified into three drainage classes.…”