The limitations of using Cs-137 for estimating soil redistribution in semi-arid environments

The Cs-137 technique for estimating net (ca. 30 years) soil flux has been u sed successfully in many environments. Its widespread use is probably becau se the Cs-137 technique overcomes many of the problems of monitoring soil e rosion and deposition (flux) over the medium-term (5 to 50 years) and at th e hillslope scale. In this respect, the technique probably offers the great est potential for measuring net soil flux in semi-arid environments where s oil flux monitoring difficulties are compounded by considerable spatial and temporal variability of the controlling factors. However, there remain unc ertainties in the underlying assumptions of the technique and difficulties in satisfying these assumptions, especially in semi-arid areas. Several key assumptions of the Cs-137 technique for measuring net soil flux in semi-ar id areas were investigated using data from southwest Niger. Samples were ob tained along a toposequence typical of the region and at nested grid nodes stratified using geomorphological information for geostatistical estimation at unsampled locations providing complete coverage of the study area. The pervasive occurrence of dust made it difficult to identify an undisturbed s ite. A first approximation of the reference Cs-137 inventory (2066 +/- 125 Bq m(-2)) was provided by modelling the Cs-137 profile at an unvegetated si te. Despite little evidence in the literature on problems of identifying a reference inventory, especially those suffering from wind erasion, it is li kely that similar problems occur in other semi-arid areas, problems with th e preferential transport of Cs-137 were identified by expressing the Cs-137 concentration as a proportion of the weight in each grain-size fraction. H owever, it was partially accounted for in the calibration relationship by r educing (by a factor of 10) the Cs-137 concentration of soil samples from t he plateau. It is likely that workers in other semi-arid regions have simil ar problems of preferential transport of Cs-137 and the method discussed he re appears to be a valuable tool to indicate the potential for preferential Cs-137 movement. A large disparity was found between the 'undisturbed' mod el and the structure of a Cs-137 profile from a small alluvial fan. This wa s due to the removal of Cs-137-rich soil and its replacement with soil larg ely devoid of Cs-137, which probably originated from gully walls. Two separ ate models that related Cs-137 movement to soil redistribution were used to calculate net soil flux. Additional modelling would be required to account for Cs-137 dilution in order to measure net soil flux in gullied and badly eroded rangeland areas. Net soil flux was calculated at sites along the to posequence, at the nested grid sites and at unsampled locations across the study using geostatistics. The toposequence samples considerably underestim ated the net soil loss relative to the nested grid samples and the geostati stical estimates of net soil flux. Toposequence sampling does not account f or the spatial variation in net soil flux unless very well-defined geomorph ological units that control the soil redistribution processes are evident. This form of sampling is of limited value for investigating the variation o f Cs-137 depth distribution at all but the smallest scale (largest areas) s ince it cannot easily be used to determine local variation in soil redistri bution processes. A small improvement in net soil flux accuracy was provide d by the geostatistical estimates, relative to the nested grid samples of n et soil flux. This was thought to be due to the efficiency of the nested grid samples sug gesting that geomorphological information can be used to limit the number o f samples necessary to encompass several scales of net soil flux variation. (C) 1999 Elsevier Science B.V. All rights reserved.