Multivariate geospatial analysis for predicting soil variability along a toposequence of a watershed in the humid tropics

Abstract

Spatiotemporal patterns of edaphic properties can be useful in modeling underlying mechanisms and controls of watershed processes, however, few studies have quantified the spatial variability of soils in watersheds in the humid tropics. A multivariate geostatistical approach was used to examine the relationships between soil apparent electrical conductivity (ECa) and measured soil and landform parameters, as well as to assess the temporal and spatial stability of ECa in a 50 × 150 m sub-watershed with three distinct land uses (forest, savannah grassland, and cultivation) in the Northern Range of Trinidad. Soil samples (0–0.2 m), and ECa measured at shallow (0 – 0.5 m; ECas) and deep (0 – 1.6 m; ECad) depths of exploration using a DUALEM-1S EC meter, were taken approximately 4 m apart along a 120 m transect in the toposequence. Although ECa and soil properties (e.g., soil depth and texture) differed among the steeply sloping forest, gently sloping savannah grassland, and flat cultivated area, there was no evidence of an effect of land use on soil properties other than what could be explained by the slope gradient. Sixty-one percent of the variability in ECas was explained by soil depth and water-stable aggregates whereas, 87% of the variation in ECad was explained by soil depth, clay content, and soil pH. Soil depth explained the greatest proportion of the variability in ECa in both models. Additionally, repeated ECa mappings revealed that ECas and ECad were spatiotemporally stable. Thus, ECad which is strongly influenced by the underlying bedrock can be used to model the spatial variability in soil depth, whereas ECas may be better for predicting surface soil properties. Models developed using this approach can predict field-scale variations of soil depth, thereby improving the accuracy of hydro-ecological models.