Although high-latitude loess sequences have been studied extensively for reconstructing past climates, loess in desert margins has been mostly overlooked. Moreover, many low-latitude loess studies have focused on secondary, reworked sequences for which paleoclimatic inferences are complicated owing to postdepositional processes acting under local control. In this respect, the spatial distribution of the loess and its different types are crucial. In this study, we differentiate primary and secondary loess using remote sensing and field criteria. We present remote sensing and geographic information, system procedures for mapping loess distribution, in general, and for identifying pristine hilltop loess sequences in particular. Then, we present the mineralogical and textural analyses combined with optically stimulated luminescence ages of one such, primary, hilltop loess sequence in the Negev desert that show that the late Pleistocene Negev loess started accumulating at ∼95 ka, much earlier than previously proposed (∼70 ka). It was preceded by a previously unknown middle Pleistocene period of loess formation at ∼180-130 ka. The ages also suggest that the threshold of dust accretion rate needed for loess formation is ∼0.02 mm a-1 (∼30 g m -2 a-1), a factor of 2 or more lower than previously thought (0.04 to 1.0 mm a-1). Our results also provide insights into causes of late Pleistocene loess accretion and its close association with regional dune field activity. The methodology for mapping the loess distribution may potentially be generalized to help understanding loess formation in other, more remote desert areas.