Item

Abstract

The Ili-Balkhash region in southeastern Kazakhstan hosts morphologically diverse dormant desert dune fields and presents an interesting opportunity for geomorphological and palaeoenvironmental studies. Because the morphology of aeolian dunes is primarily driven by wind dynamics, the dormant dunes in the study area may reflect past wind conditions. We assess their concurrence with modern ERA5 wind data to test whether there has been a change in wind regime since the dunes' last phase of activity. Our approach includes dune mapping, the quantification of dune orientations, the modelling of modern bedform orientations, and optically stimulated luminescence (OSL) dating for temporal context. The centrepiece of our methodological contribution is a novel semi-automated mapping workflow using geographic object-based image analysis (GEOBIA) and machine learning (ML) on Sentinel-2 satellite imagery. Within the scope of a case study, we map dune fields in the Ili-Balkhash region and quantify dune orientations. We further apply the maximum gross bedform-normal transport (MGBNT) concept to model bedform orientations matching modern wind regimes for each of the sites. We find that strong winds show better alignment with observed dune orientations than wind regimes comprising all wind speeds. Furthermore, bedform orientations in some of our study sites, namely those that are located in the open plain southeast of Lake Balkhash, do not reflect modern winds. The divergence between dune orientations and wind regime suggests changes in local wind dynamics since the dune fields' last phase of activity.