Spatial variability of soil aggregate stability at the scale of an agricultural region in Tunisia

Abstract

Soil aggregate stability is a key factor in soil resistance to water erosion, which is a threat to soils in a large part of northern Tunisia. The analysis of the spatial variability of soil aggregate stability provides both agronomic and environmentally useful information. However, extensive measurements of soil aggregate stability remain tedious and expensive.This study explores two different approaches as alternative to measurements of soil aggregate stability. One approach estimated aggregate stability via laboratory measurements of soil elementary properties using multiple linear regressions known as pedotransfer functions. The second approach, which is methodologically innovating, was based on the geological pattern as a proxy for aggregate stability using regression-kriging analysis. A set of 113 soil samples from an 800km2 agricultural region that included the Lebna watershed (Cap Bon, Tunisia) were collected from the soil surface layer (0–10cm depth). Samples were analyzed for elementary properties (i.e., soil texture, total carbon and nitrogen, iron, CaCO3, salinity, CEC and pH) and for soil aggregate stability according to the normalized method (ISO/DIS 10930, 2012), which considers three indexes (MWD) calculated for three contrasted wetting conditions and disruptive energies.Most soils in the study area were non-salted with an alkaline pH and relatively low organic carbon content. Of the soils, 35% were clay soils, and 55% had a balanced soil texture. The average of the three soil aggregate stability indexes (MWDmean) ranged from 0.38 to 2.80mm, and this property showed large variability from instable soils to very stable ones. Analysis of pedotransfer functions determined that the best predictor variables for soil aggregate stability were silt, organic matter and iron. Geostatistical analyses at the regional scale showed spatially structured soil aggregate stability (variograms with sills reaching a 5km distance). Using geological information as ancillary data, the prediction of soil aggregate stability with regression-kriging was similar to that of pedotransfer functions. A regression-kriged map of soil aggregate stability associated with a map of prediction uncertainties was developed. The resulting maps and methods of this study can be useful in the development of management options that minimize water erosion risks in the studied area.