Modeling Soil-Landscape Relationships

• Main Author: Perpustakaan UGM, i-lib
• Format: Article NonPeerReviewed
• Published: [Yogyakarta] : Universitas Gadjah Mada 2005
• Subjects: Jurnal i-lib UGM
• Online Access: https://repository.ugm.ac.id/27265/; http://i-lib.ugm.ac.id/jurnal/download.php?dataId=10317
• Institution: Universitas Gadjah Mada

Soil-landscape model represents both relationships between soil and landform and the relationships between the pattern of soil-landform relationship and processes of pedogeomorphic evolution. This paper aims to get insight about soil-landscape relationships as a basis for soil-landscape modeling with giving emphasis on the methodology and the review of the modeling result of research from developed countries since the advent of GIS. The modeling follows four consecutive stages: physiographic domain characterization, geomorphometric characterization of the landscape, horizon stratigraphy characterization and soil property characterization. Landscape position determines soil profile properties and horizon properties. At lar~e scale, slope, flow accumulation, and CTI are best predictors for A-horizon depth (R = 0.85), whereas CTI alone accounts for 71 % of variance. Flow accumulation and upslope mean profile curvature are best predictors for soil depth (R2= 0.88), whereas CTI alone accounts for 84 % of variance. Slope and flow accumulation are best predictor for carbon content (R2 = 0.80), and CTI accounts for 78 % of variance. A tree with five terminal nodes was optimal for predicting soil profile depth with slope, CTI, relative elevation, and temperature as predictors. A tree with five terminal nodes was optimal for predicting total soil phosphorus with radiometric potassium, relief, downslope gradient, and plan curvature as predictors. At intermediate to small scale, gilgai and landform are best ~redictor for clay content (R2 = 0.51). Gilgai is best predictor for prediction of EC (R =0.364), and landform is best predictor for ESP (R2=0.561). A tree with six terminal nods was optimal for predicting depth of A horizon of tuff soil in Lampung Province, Indonesia with average slope and elevation as predictors. Modeling has increasingly developed mainly due to the advance of computer technology both software and hardware. Generalized linear model and tree-based model can be used to develop quantitative soil-landscape relationship Keyword: Soil-landscape relationship, modeling, regression tree, environmental correlation, Lampung